Method and device for automatically testing sensitivity of broadband radio frequency system and medium

Abstract

The invention discloses a method, equipment and medium for automatically testing the sensitivity of a broadband radio frequency system, and relates to the technical field of broadband radio frequency testing, and the method comprises the steps: starting a testing process, and coordinating a plurality of calibration equipment and equipment to be tested in a wireless network in the same testing time slot, synchronously sampling spectrum data of each calibration device through an air interface and processing the spectrum data into a noise distribution parameter set; and dynamically adjusting the output power of the vector signal generator and synchronously monitoring the bit error rate change according to the received power predicted value until the difference value between the bit error rate and a preset target bit error rate threshold value meets a convergence condition, recording the received power value during convergence as a sensitivity test result, and generating a test report. By generating the high-precision dynamic noise map and calculating the compensation transmitting power calibration value, accurate sensing and active compensation of the space noise of the test environment can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wideband radio frequency testing, and in particular to a method and device for automatically testing the sensitivity of a wideband radio frequency system, and a medium. BACKGROUND

[0002] The testing of the sensitivity of a wireless communication device is a key link for ensuring the quality and reliability of communication. The conventional testing method is usually carried out in an anechoic chamber or a static environment, and uses a vector signal generator and a spectrum analyzer and other devices to determine the sensitivity threshold by manually adjusting the transmission power point by point and observing the bit error rate change. With the development of the fifth generation mobile communication technology (5G) towards millimeter wave and large-scale antenna array, the requirements for testing efficiency and accuracy are increasing, and automatic testing has gradually become an industry standard configuration.

[0003] However, most of the existing automatic testing schemes are based on the assumption of ideal static noise environment, and there is a difference between the test environment construction and the complex electromagnetic environment in the real commercial scene. In particular, in the dynamic networking test scene of multiple devices and multiple frequency bands coexisting, the background noise is time-varying and the spatial distribution is uneven, and the traditional method is difficult to realize real-time sensing and compensation for the interference of spatial noise fluctuation on the test result, resulting in the decrease of sensitivity test accuracy, the increase of test cycle times and the low overall efficiency, which cannot meet the demand of modern radio frequency rapid and accurate testing. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a method for automatically testing the sensitivity of a wideband radio frequency system to solve the problems of insufficient radio frequency receiving sensitivity test accuracy and low efficiency under complex dynamic noise environment.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a method for automatically testing the sensitivity of a wideband radio frequency system, which comprises: starting a test process; coordinating multiple calibration devices and a device under test in a wireless network to synchronously sample the spectrum data of the calibration devices and process the spectrum data into a set of noise distribution parameters in the same test time slot; uploading the set of noise distribution parameters to a master control end for fusion to generate a high-precision dynamic noise map; extracting the noise power value of the location of the device under test from the high-precision dynamic noise map and obtaining a compensation transmit power calibration value; adjusting the output power of a vector signal generator according to the compensation transmit power calibration value to generate a wideband test signal and transmit the wideband test signal to the device under test, and reading the internal state information of the receiving link of the device under test in real time; inputting the internal state information into a machine learning model to output a predicted value of the receiving power required to reach a target bit error rate under the current test condition; dynamically adjusting the output power of the vector signal generator according to the predicted value of the receiving power and synchronously monitoring the change of the bit error rate until the difference between the bit error rate and a preset target bit error rate threshold meets a convergence condition, recording the receiving power value at the time of convergence as the sensitivity test result and generating a test report.

[0007] As a preferred scheme of the method for automatically testing the sensitivity of the wideband radio frequency system, the spectrum data of the calibration devices is synchronously sampled through the air interface and processed into a set of noise distribution parameters, and the specific steps are as follows, The master control end starts the test process, collects the spatial position coordinates of the calibration devices, and establishes nanosecond-level time synchronization with the calibration devices and the device under test through a precise time protocol to allocate the same sampling window. The calibration devices perform radio frequency sampling using the local oscillator frequency and the sampling rate in the same sampling window to obtain sampling data. The sampling data is windowed and processed to generate the spectrum data of the calibration devices through fast Fourier transform. The spectrum data of the calibration devices is fitted with a Gaussian distribution to extract noise mean parameters and noise variance parameters. The spatial position coordinates of the calibration devices, the noise mean parameters, and the noise variance parameters are combined to generate a set of noise distribution parameters.

[0008] As a preferred scheme of the method for automatically testing the sensitivity of the wideband radio frequency system, the high-precision dynamic noise map is generated, and the specific steps are as follows, The master control end receives the set of noise distribution parameters and performs standardization preprocessing to generate a set of standardized noise distribution parameters. Based on the set of standardized noise distribution parameters, the Kriging spatial interpolation algorithm is used to calculate the noise mean estimate value and the noise variance estimate value of any position in the test area to form initial noise distribution data. The initial noise distribution data is input into a time Kalman filter for time dimension filtering and optimization to generate optimized noise distribution data; The optimized noise distribution data is subjected to multi-resolution fusion processing to generate a high-precision dynamic noise map.

[0009] As a preferred scheme of the method for automatically testing the sensitivity of the wideband radio frequency system, the specific steps for calculating the compensation transmit power calibration value are as follows, The spatial position coordinates of the device under test are acquired, and the high-precision dynamic noise map is queried to obtain the noise mean value estimate and the noise variance estimate of the position where the device under test is located; Based on the noise mean value estimate and the noise variance estimate, the noise power value is calculated using a noise power synthesis algorithm; Taking the spatial position coordinates of the device under test as the center, spatial data points and noise parameters within a preset neighborhood range are extracted from the high-precision dynamic noise map to form a local noise parameter set; According to the local noise parameter set, the spatial uncertainty weight factor is calculated, and the noise power value is corrected to obtain an initial compensation power calibration value; The initial compensation power calibration value is subjected to amplitude limiting and time sequence smoothing processing to obtain the compensation transmit power calibration value.

[0010] As a preferred scheme of the method for automatically testing the sensitivity of the wideband radio frequency system, the specific steps for real-time reading of the internal state information of the device under test are as follows, The compensation transmit power calibration value is converted through the SCPI protocol format to generate an instrument control instruction; The vector signal generator adjusts the output power according to the instrument control instruction, and generates a wideband test signal through direct digital frequency synthesis and I / Q modulation algorithms; The device under test demodulates the wideband test signal through the internal receiving chain to generate internal state information, which is read in real time through the diagnostic interface.

[0011] As a preferred scheme of the method for automatically testing the sensitivity of the wideband radio frequency system, the internal state information includes automatic gain control voltage, channel estimation coefficient, and equalizer tap weight.

[0012] As a preferred scheme of the method for automatically testing the sensitivity of the wideband radio frequency system, the specific steps for outputting the receive power prediction value required to reach the target bit error rate under the current test condition are as follows, The automatic gain control voltage, channel estimation coefficient, and equalizer tap weight in the internal state information are combined with historical state data to construct a space-time joint feature tensor; The spatio-temporal joint feature tensor is input into a machine learning model to extract a deep spatio-temporal feature vector; According to the deep spatio-temporal feature vector, a posterior probability distribution of the received power prediction value is calculated by Bayesian optimization inference, and the received power prediction value required to reach the target bit error rate is output.

[0013] As a preferred scheme of the method for automatically testing the sensitivity of the wideband radio frequency system, the received power value at convergence is recorded as the sensitivity test result and a test report is generated, and the specific steps are as follows, The initial output power of the vector signal generator is set based on the received power prediction value; The bit error rate data is monitored in real time, and a power control strategy is dynamically generated according to the difference between the bit error rate data and the preset target bit error rate threshold, and the output power of the vector signal generator is adjusted; According to the adjusted output power, the bit error rate change of the device under test is monitored synchronously, and the bit error rate difference between the new bit error rate data of the device under test and the preset target bit error rate threshold is calculated; It is judged whether the bit error rate difference meets the preset convergence determination parameter, and when it meets, the current received power value is recorded as the sensitivity test result; According to the sensitivity test result and the data of the entire test process, a test report is automatically generated.

[0014] In a second aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program is executed by the processor to implement any step of the method for automatically testing the sensitivity of the wideband radio frequency system according to the first aspect of the present application.

[0015] In a third aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement any step of the method for automatically testing the sensitivity of the wideband radio frequency system according to the first aspect of the present application.

[0016] The present application has the following advantages: by generating a high-precision dynamic noise map and calculating a compensation transmit power calibration value, accurate perception and active compensation of the spatial noise of the test environment can be achieved. By fusing multi-source noise data to construct a visual electromagnetic environment model, the power pre-adjustment of the transmit signal has environmental perception capability, so that the received power of the device under test can be quickly approximated to the vicinity of the real sensitivity threshold in the initial stage of the test, reducing the number of iterations and time required for fine power adjustment. The test error caused by environmental noise fluctuation is effectively suppressed, the accuracy and reliability of the sensitivity test result are improved, a solid foundation is laid for the accelerated convergence of the entire automated test process, and the test efficiency is improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Fig. 1 Flow chart of the method for automatic test of sensitivity of a wideband radio frequency system.

[0019] Fig. 2 Flow chart for generating a set of noise distribution parameters.

[0020] Fig. 3 Flow chart for generating a high-precision dynamic noise map.

[0021] Fig. 4 Flow chart of the closed-loop control for sensitivity test. DETAILED DESCRIPTION

[0022] In order to make the above objectives, features and advantages of the present application more apparent and comprehensible, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, so the present application is not limited to the specific embodiments disclosed below.

[0024] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0025] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a method for automatic test of sensitivity of a wideband radio frequency system, comprising the following steps: S1, starting the test process, coordinating multiple calibration devices and devices to be tested in the wireless networking to synchronously sample the spectrum data of each calibration device through the air interface in the same test time slot and processing into a set of noise distribution parameters.

[0026] S1.1: the master control end starts the test process, collects the spatial position coordinates of each calibration device, and establishes nanosecond-level time synchronization with each calibration device and the device to be tested through the precision time protocol, and distributes the sampling window of the same time slot; Specifically, the master end sends a start command, initializes the test process, triggers each device to enter a preparation state, obtains the spatial position coordinates of each calibration device through indoor positioning, records the three-dimensional coordinate values of each calibration device, including east coordinate, north coordinate and height coordinate; The precise time protocol is used to synchronize time with all calibration devices and devices to be tested, clock offset and delay are obtained through exchange of timing messages, local clocks of each device are adjusted, and nanosecond-level time synchronization is realized. According to the synchronized time, a fixed same time slot is allocated as a sampling window, the start time and duration of the sampling window are specified, and each device is notified.

[0027] S1.2: Each calibration device uses a local oscillator frequency and a sampling rate to perform radio frequency sampling in the sampling window of the same time slot to obtain sampling data. Specifically, after each calibration device receives the sampling window instruction from the master end, the radio frequency receiving chain is started at the specified start time. The calibration device uses a preset local oscillator frequency and sampling rate to perform radio frequency sampling. The local oscillator frequency is generated by a local oscillator, and the sampling rate is set by an analog-to-digital converter. During radio frequency sampling, in-phase and quadrature components are collected to generate analog radio frequency signals, which are converted to sampling data by down-conversion through a local oscillator and sampling conversion through an analog-to-digital converter.

[0028] It should be noted that the preset local oscillator frequency is determined according to the center frequency of the radio frequency band range supported by the device to be tested, and the example value is 2.4GHz, which ensures that the sampling covers the working range of the target frequency band; The preset sampling rate needs to meet the Nyquist sampling theorem, i.e. the sampling rate is greater than or equal to twice the signal bandwidth. For example, if the test signal bandwidth is 100MHz, the sampling rate needs to be at least 200MSa / s to avoid signal distortion caused by spectral aliasing.

[0029] S1.3: Window processing is performed on the sampling data, and frequency spectrum data of each calibration device is generated through fast Fourier transform; Specifically, a window function is applied to the sampling data, and the window function is selected from a Hanning window, a Hamming window or a Blackman window. Fast Fourier transform is performed on the windowed sampling data, the number of points of the fast Fourier transform matches the length of the sampling data, and the fast Fourier transform directly generates a frequency domain complex spectrum. The frequency domain complex spectrum is converted to an amplitude spectrum, which is used as the frequency spectrum data of each calibration device.

[0030] S1.4: Gaussian distribution fitting is performed on the frequency spectrum data of each calibration device to extract noise mean parameter and noise variance parameter; Specifically, the amplitude value of each frequency point in the spectrum data of each calibration device is converted into a corresponding spectrum data power value, the power values of the spectrum data of each calibration device are taken as samples, a Gaussian distribution fitting is performed by using a maximum likelihood estimation method, and noise mean parameters and noise variance parameters representing the noise power distribution are directly obtained through the fitting process. The noise mean parameters represent the concentration trend of the spectrum data power values, and the noise variance parameters represent the dispersion degree of the spectrum data power values.

[0031] S1.5: The spatial position coordinates, noise mean parameters and noise variance parameters of each calibration device are combined to generate a noise distribution parameter set.

[0032] Specifically, each calibration device uses its own spatial position coordinates, noise mean parameters and noise variance parameters to pack into a data packet, and the data packet of each calibration device is transmitted to the host end through wireless or wired communication. The host end receives the data packets of all calibration devices, and combines the data packets of all calibration devices into a noise distribution parameter set.

[0033] S2, upload the noise distribution parameter set to the host end for fusion to generate a high-precision dynamic noise map.

[0034] S2.1: The host end receives the noise distribution parameter set and performs standardization preprocessing to generate a standardized noise distribution parameter set; Specifically, the host end receives the noise distribution parameter set through the communication link, determines the arithmetic mean of all noise mean parameters as a mean reference benchmark, determines the standard deviation of all noise mean parameters as a mean dispersion degree measure, the difference between the noise mean parameter of each calibration device and the mean reference benchmark as an intermediate mean reference benchmark, and the ratio between the intermediate mean reference benchmark and the mean dispersion degree measure as a standardized noise mean parameter. determines the arithmetic mean of all variance mean parameters as a variance reference benchmark, determines the standard deviation of all noise variance parameters as a variance dispersion degree measure, the difference between the noise variance parameter of each calibration device and the variance reference benchmark as an intermediate variance reference benchmark, and the ratio between the intermediate variance reference benchmark and the variance dispersion degree measure as a standardized noise variance parameter; The spatial position coordinates remain unchanged, the standardized noise mean parameters, the standardized noise variance parameters and the spatial position coordinates of all calibration devices are combined to form a standardized noise distribution parameter set.

[0035] S2.2: Based on the standardized noise distribution parameter set, the Kriging spatial interpolation algorithm is used to calculate the noise mean estimate value and the noise variance estimate value of any position in the test area to form an initial noise distribution data. Specifically, interpolation is performed on all grid points in the test area, and the noise mean estimate and the noise variance estimate in the test area after interpolation are integrated to form initial noise distribution data.

[0036] The noise mean estimate and the noise variance estimate at any position in the test area are calculated using the Kriging spatial interpolation algorithm, and the expression is: ; ; In the formula, represents the noise mean estimate, represents the target position coordinates, represents the total number of calibration devices, represents the calibration device index, represents the interpolation weight of the th calibration device for the target position coordinates, represents the normalized noise mean parameter, represents the position coordinates of the th calibration device, represents the noise variance estimate, represents the normalized noise variance parameter.

[0037] It should be noted that , and are dimensionless, and the outputs and are dimensionless, keeping the dimensions consistent.

[0038] S2.3: Input the initial noise distribution data into the time Kalman filter for time dimension filtering and optimization to generate optimized noise distribution data; Specifically, the initial noise distribution data is configured as the state vector of the time Kalman filter, and the state vector is composed of the noise mean estimate and the noise variance estimate of each grid point arranged in order; The time Kalman filter recursively performs state prediction operation and state update operation according to discrete time steps. The state prediction operation is based on the previous state vector and the state transition matrix to generate the state prediction value at the current time. The state update operation fuses the state prediction value at the current time with the corresponding value in the initial noise distribution data at the current time to generate the posterior state estimate. The posterior state estimate of all grid points constitutes the optimized noise distribution data.

[0039] S2.4: Perform multi-resolution fusion processing on the optimized noise distribution data to generate a high-precision dynamic noise map.

[0040] Specifically, the optimized noise distribution data is divided according to a spatial resolution level, and the original grid point noise mean value and noise variance value are retained to form a base layer. The base layer grid points are merged according to adjacent rows and columns to form a resolution halved middle layer. The noise mean value of the middle layer is obtained by averaging the original values of the merged regions, and the noise variance estimation value is obtained by selecting the maximum value of the original values of the merged regions. The middle layer is merged according to adjacent rows and columns to form a resolution halved high layer. A tower fusion algorithm is used to perform hierarchical fusion. The high layer is expanded to the middle layer resolution to form a virtual middle layer by bilinear interpolation. The middle layer and the virtual middle layer are superimposed at the corresponding grid points to form a fused middle layer. The fused middle layer is expanded to the base layer resolution to form a virtual base layer by bilinear interpolation. The base layer and the virtual base layer are superimposed at the corresponding grid points to form a fused base layer. The fused base layer is used as a high-precision dynamic noise map.

[0041] S3. Extracting the noise power value of the position of the device to be tested from the high-precision dynamic noise map, and obtaining the compensation transmit power calibration value.

[0042] S3.1: Collecting the spatial position coordinates of the device to be tested, querying the high-precision dynamic noise map, and obtaining the noise mean value estimation and noise variance estimation of the position of the device to be tested; Specifically, the spatial position coordinates of the device to be tested are obtained by indoor positioning. The coordinate system of the spatial position coordinates of the device to be tested is completely consistent with the coordinate system used by the calibration device. The spatial position coordinates of the device to be tested are matched with the grid point positions of the high-precision dynamic noise map. The nearest grid point to the spatial position coordinates of the device to be tested in the high-precision dynamic noise map is located. The noise mean value estimation and noise variance estimation are directly read from the nearest grid point.

[0043] S3.2: Based on the noise mean value estimation and the noise variance estimation, the noise power value is calculated by using a noise power synthesis algorithm; ; In the formula, represents the noise power value, represents a physical unit conversion constant, represents 1 / 10 scale of the noise mean value estimation, represents conversion of to milliwatt unit, represents a fluctuation contribution adjustment coefficient, represents 1 / 10 scale of the noise variance estimation, represents conversion of to milliwatt unit.

[0044] It should be noted that the fluctuation contribution adjustment coefficient is set according to the noise fluctuation characteristics of the test scene. The example value is 0.3.

[0045] S3.3: Extracting spatial data points and noise parameters within a preset neighborhood range from the high-precision dynamic noise map centered on the spatial position coordinates of the device under test, to form a local noise parameter set; Specifically, taking the collected spatial position coordinates of the device under test as the center and the radius of the preset neighborhood range defining a circular area, all grid points of the high-precision dynamic noise map are traversed, the Euclidean distance between the spatial position coordinates of each grid point and the spatial position coordinates of the device under test is calculated by the Euclidean distance formula, and the grid points whose Euclidean distance does not exceed the radius of the preset neighborhood range are screened. The spatial position coordinates, noise mean estimate value and noise variance estimate value of the grid points that do not exceed the radius of the preset neighborhood range are read from the high-precision dynamic noise map. The spatial position coordinates, noise mean estimate value and noise variance estimate value of each grid point constitute a grid point data packet, and the grid point data packets of all screened grid points are combined to form a local noise parameter set.

[0046] It should be noted that the preset neighborhood range refers to a spherical spatial area with the spatial position coordinates of the device under test as the center and the radius of the preset neighborhood range as the radius. The preset neighborhood range is set based on the spatial resolution of the high-precision dynamic noise map, and the radius of the preset neighborhood range is set based on the physical size of the device under test and the spatial correlation of the test scene. The example value is 0.5 meters to 2 meters.

[0047] S3.4: According to the local noise parameter set, the spatial uncertainty weight factor is calculated, and the noise power value is corrected to obtain the initial compensation power calibration value; Specifically, the noise variance estimate values of all grid points in the local noise parameter set are weighted and averaged by the spatial uncertainty weight factor to generate a weighted average noise variance. The product of the preset spatial uncertainty correction coefficient and the weighted average noise variance is used as the spatial uncertainty correction amount, and the spatial uncertainty correction amount is added to the noise power value to generate the initial compensation power calibration value.

[0048] It should be noted that the preset spatial uncertainty correction coefficient is set according to the spatial interpolation error characteristics of the high-precision dynamic noise map and the sensitivity of the receiving link of the device under test. The example value is 0.1.

[0049] According to the local noise parameter set, the spatial uncertainty weight factor is calculated, and the expression is: ; In the formula, represents the spatial uncertainty weight factor of the th neighborhood grid point, represents the index of the neighborhood grid point, represents the spatial position coordinates of the device under test, represents the spatial position coordinates of the th neighborhood grid point, This represents a minimum positive number excluding zero.

[0050] It should be noted that, , and All are dimensionless, and the final output is... It is dimensionless, but we maintain dimensional consistency.

[0051] S3.5: Perform amplitude limiting and timing smoothing on the initial compensated power calibration value to obtain the compensated transmit power calibration value.

[0052] Specifically, a minimum compensation power calibration value threshold and a maximum compensation power calibration value threshold are set, and the relationship between the initial compensation power calibration value and the minimum compensation power calibration value threshold and the maximum compensation power calibration value threshold is compared. If the initial compensation power calibration value is less than the minimum compensation power calibration value threshold, then the initial compensation power calibration value is set to the minimum compensation power calibration value threshold. If the initial compensation power calibration value is greater than the maximum compensation power calibration value threshold, then the initial compensation power calibration value is set to the maximum compensation power calibration value threshold. Otherwise, the initial compensation power calibration value remains unchanged. The compensated power calibration value after amplitude limiting is processed by performing time-series smoothing. The compensated transmit power calibration value of the previous moment is obtained. The current compensated power calibration value after amplitude limiting is weighted and fused with the compensated transmit power calibration value of the previous moment. The weighting and fusion weight adopts a preset smoothing factor to generate the compensated transmit power calibration value.

[0053] It should be noted that the minimum compensation power calibration threshold is set based on the minimum detectable power of the receiving link of the device under test (DUT), with an example value of -90, which is the lower limit of typical receiver sensitivity, ensuring that the signal is higher than the noise floor. The maximum compensation power calibration threshold is set based on the maximum input power tolerance of the receiving link of the DUT, with an example value of -30, which is the maximum input power limit of common RF front-ends (such as the saturation point of LNA / ADC), protecting the RF front-end from saturation. The preset smoothing factor is set based on the time-varying fluctuation period of the test environment noise, with an example value of 0.2. Under typical noise fluctuation periods (such as hundreds of milliseconds), a weight of 0.2 can effectively suppress short-term fluctuations (such as sudden interference) while preserving long-term trends (such as device temperature drift), balancing response speed and stability.

[0054] S4. Adjust the output power of the vector signal generator according to the compensation transmit power calibration value, generate a broadband test signal and transmit it to the device under test, and read the internal status information of the receiving link of the device under test in real time.

[0055] S4.1: Convert the compensated transmit power calibration value through the SCPI protocol format to generate instrument control commands; Specifically, the power setting command word of the SCPI protocol is used as the instruction prefix, the power unit symbol of the SCPI protocol is used as the instruction suffix, the numerical part of the compensation transmit power calibration value, the power setting command word and the power unit symbol are combined into a complete instruction string as the instrument control instruction.

[0056] S4.2: The vector signal generator adjusts the output power according to the instrument control instruction, and generates a wideband test signal through direct digital frequency synthesis and I / Q modulation algorithm. Specifically, the vector signal generator receives and analyzes the instrument control instruction, adjusts the gain of the internal power amplifier according to the power setting value in the instrument control instruction, steps the phase sequence through the phase accumulator of the direct digital frequency synthesis according to the frequency control word, the phase sequence is addressed in real time through the preset sine / cosine lookup table to generate the in-phase component and the quadrature component, the in-phase component and the quadrature component are modulated to the radio frequency carrier through the I / Q modulation algorithm, the in-phase component is mixed with the local oscillator cosine signal, the quadrature component is mixed with the local oscillator sine signal, the mixing results are added to generate the radio frequency signal, and the radio frequency signal is amplified by the power amplifier to output the wideband test signal.

[0057] It should be noted that the preset sine / cosine lookup table is a discrete sine / cosine function value table pre-stored in the ROM (such as mapping 0~360° phase to 4096 quantization values), and the phase sequence is used as the address of the lookup table to directly index the corresponding in-phase component and quadrature component in the sine / cosine lookup table.

[0058] S4.3: The device under test demodulates the wideband test signal through the internal receiving link, generates internal state information, and reads it in real time through the diagnostic interface.

[0059] Specifically, the internal receiving link of the device under test receives the wideband test signal, adjusts the signal amplitude of the wideband test signal through the automatic gain control circuit, generates a radio frequency signal, inputs the radio frequency signal into a mixer, generates an intermediate frequency signal through frequency down-conversion, suppresses high frequency components through an anti-aliasing filter, generates a filtered intermediate frequency signal, inputs the filtered intermediate frequency signal into an analog-to-digital converter, performs sampling and holding and quantization coding at a preset sampling rate, generates a digital intermediate frequency signal sequence, and generates a baseband in-phase signal and a baseband quadrature signal through quadrature demodulation. The digital intermediate frequency signal sequence is subjected to carrier synchronization and channel response recovery to eliminate phase shift, and the synchronous baseband in-phase signal and the synchronous baseband quadrature signal are generated. The synchronous baseband in-phase signal and the synchronous baseband quadrature signal are subjected to signal demodulation to generate internal state information, and the device under test reads it continuously at the same time interval through the hardware debugging protocol interface.

[0060] S4.4: The internal state information includes the automatic gain control voltage, the channel estimation coefficient and the equalizer tap weight.

[0061] Specifically, in the signal demodulation process, the automatic gain control circuit generates an automatic gain control voltage, the channel estimation operation generates a channel estimation coefficient, and the adaptive equalizer generates an equalizer tap weight. It should be noted that the automatic gain control voltage refers to the analog voltage value output by the automatic gain control circuit, which is used to adjust the amplification factor of the variable gain amplifier to stabilize the received signal amplitude. The channel estimation coefficient is a complex array representing the frequency response characteristics of the wireless channel, generated by the least squares estimation algorithm, and used to compensate for signal phase and amplitude distortion. The equalizer tap weight refers to the real coefficient sequence of the finite impulse response filter in the time domain equalizer, which is used to suppress inter-symbol interference, and the length is determined by the device specification.

[0062] S5, input the internal state information into the machine learning model, and output the predicted value of the received power required to reach the target bit error rate under the current test condition.

[0063] S5.1: Combine the automatic gain control voltage, channel estimation coefficient and equalizer tap weight in the internal state information with the historical state data to construct a spatio-temporal joint feature tensor; Specifically, the automatic gain control voltage is expanded into a fixed-length vector, the channel estimation coefficient is decomposed into a real part vector and an imaginary part vector, and the zero padding operation is performed on the channel estimation coefficient and the equalizer tap weight with different lengths to align the length. Arrange in time frame order, each time frame contains three channel data, channel one is the expanded automatic gain control voltage vector, channel two is the channel estimation coefficient real part vector, channel three is the channel estimation coefficient imaginary part vector, and channel four is the equalizer tap weight vector. All time frames are stacked to form a spatio-temporal joint feature tensor.

[0064] S5.2: Input the spatio-temporal joint feature tensor into the machine learning model to extract deep spatio-temporal feature vectors; It should be noted that the pre-training process of the machine learning model is as follows: collect historical spatio-temporal joint feature tensors as samples and corresponding historical received power measured values, construct a three-dimensional convolutional neural network architecture, which includes an alternating sequence of three-dimensional convolutional layers, nonlinear activation layers and three-dimensional maximum pooling layers, and is connected to a fully connected layer at the end. The loss function is defined as the mean square error of the received power prediction value and the historical received power measured value, the gradient descent algorithm is used to iteratively optimize the network parameters, the gradient of the loss function with respect to the network parameters is calculated, the network parameters are updated in the opposite direction of the gradient, and the iterative process is repeated until the loss function converges. Save the network parameters to obtain the pre-trained machine learning model.

[0065] Specifically, the spatio-temporal joint feature tensor is input into the machine learning model, the machine learning model performs a three-dimensional convolution operation on the spatio-temporal joint feature tensor using a three-dimensional convolution kernel, the three-dimensional convolution result is converted by a nonlinear activation function to generate a primary feature map, the primary feature map is reduced in spatial size by a three-dimensional maximum pooling operation, and the three-dimensional convolution operation, the nonlinear activation function conversion and the three-dimensional maximum pooling operation are repeated until the time dimension is reduced to a single frame and the spatial size reaches a predefined basic size, a three-dimensional feature structure is generated, the three-dimensional feature structure is flattened into a single-dimensional high-dimensional feature sequence, and the single-dimensional high-dimensional feature sequence is linearly transformed by a fully connected layer to generate a deep spatio-temporal feature vector.

[0066] It should be noted that the predefined basic size is set based on the network architecture of the machine learning model, and an example value is: 256 channels x 4 height x 4 width.

[0067] S5.3: According to the deep spatio-temporal feature vector, the posterior probability distribution of the received power prediction value is calculated by Bayesian optimization inference, and the received power prediction value required to reach the target bit error rate is output.

[0068] Specifically, the deep spatio-temporal feature vector is used to query the historical data set of Bayesian optimization inference, the historical data set contains historical deep spatio-temporal feature vectors and corresponding historical received power measured values, the similarity between the deep spatio-temporal feature vector and the historical deep spatio-temporal feature vector is compared using a radial basis function kernel, and the posterior probability distribution mean value is generated by weighted fusion of the historical received power measured values according to the similarity; the historical deep spatio-temporal feature vector and the historical received power measured value are used as autocorrelation values, and the similarity between the deep spatio-temporal feature vector and the historical deep spatio-temporal feature vector is used as cross-correlation values, and the posterior probability distribution variance is generated using the autocorrelation values and the cross-correlation values, the posterior probability distribution is subject to a normal distribution with the posterior probability distribution mean value as the central tendency and the posterior probability distribution variance as the uncertainty measure, and the expected improvement maximum point of the posterior probability distribution is solved with the target bit error rate as the constraint condition, and the received power value corresponding to the expected improvement maximum point is taken as the received power prediction value required to reach the target bit error rate.

[0069] ; In the formula, P(x) represents a probability distribution function, P(x) represents a received power prediction value, P(x) represents a deep spatio-temporal feature vector, P(x) represents a posterior probability distribution of a received power prediction value, P(x) represents that the posterior probability distribution is subject to a normal distribution, P(x) represents a central tendency of a posterior probability distribution, P(x) represents an uncertainty measure of a posterior probability distribution.

[0070] S6. According to the received power prediction value, the output power of the vector signal generator is dynamically adjusted and the bit error rate change is synchronously monitored until the difference between the bit error rate and the preset target bit error rate threshold value meets the convergence condition, the received power value at the time of convergence is recorded as the sensitivity test result, and a test report is generated.

[0071] S6.1: Set the initial output power of the vector signal generator based on the received power prediction value; Specifically, the received power prediction value is taken as the power setting value, and the power setting value is taken as the initial output power of the vector signal generator.

[0072] S6.2: Real-time monitoring of bit error rate data, dynamically generating power control strategy and adjusting the output power of the vector signal generator according to the difference between the bit error rate data and the preset target bit error rate threshold value; Specifically, the to-be-tested device obtains real-time bit error rate data by counting the number of error bits and the total number of transmission bits per unit time through the physical layer baseband, and takes the difference between the bit error rate data and the preset target bit error rate threshold value as the bit error rate difference absolute value. If the bit error rate difference absolute value is greater than the preset first threshold value, a fixed step power adjustment strategy is generated, if the bit error rate difference absolute value is between the preset first threshold value and the preset second threshold value, a proportional adjustment power adjustment strategy is generated, and if the bit error rate difference absolute value is less than the preset second threshold value, a fine-tuning power adjustment strategy is generated. After each strategy is generated, a new power instruction string is generated in accordance with the SCPI protocol format, and the vector signal generator executes the new power instruction string to adjust the output power.

[0073] It should be noted that the preset target bit error rate threshold value is set based on the communication standard specification of the to-be-tested device, and an example value is 0.000001. The value of 0.000001 is the bit error rate threshold of typical high-reliability communication (such as WiFi 6 / 5G) (such as the theoretical value when SNR≥15dB under QPSK modulation), and a value lower than 0.000001 will affect the data transmission integrity (such as video stream stuttering and TCP retransmission surge); the preset first threshold value is set based on the order of magnitude of the target bit error rate, and an example value is 0.001. When the bit error rate deteriorates to 0.001 (such as SNR<8dB), it is in a near-outage state (BER sharp rise curve inflection point), and the power needs to be adjusted significantly (fixed step) to quickly escape the dangerous area; the preset second threshold value is set based on the convergence accuracy requirement, and an example value is 0.00001. If the second threshold value is too small (such as <0.000001), the power fine-tuning is easy to be disturbed by noise and cause oscillation, and if it is too large (such as >0.0001), the convergence speed is slow.

[0074] The fixed step power adjustment strategy refers to the adjustment of the output power of the vector signal generator by a fixed change amount (such as ±1 decibel milliwatt); The proportional power adjustment strategy refers to a linear proportional relationship between the output power change of the vector signal generator and the absolute value of the difference between the error rate and the target error rate threshold. The fine-tuning power adjustment strategy refers to adjusting the output power of the vector signal generator by a small change (e.g., ±0.1 dBm).

[0075] S6.3: According to the adjusted output power, the error rate change of the device under test is monitored synchronously, and the error rate difference between the new error rate data of the device under test and the preset target error rate threshold is calculated. Specifically, after the vector signal generator completes the output power adjustment, the device under test re-counts the number of error bits and the total number of transmission bits per unit time after a preset stabilization time, generates a new real-time error rate, and calculates the difference between the new real-time error rate and the preset target error rate threshold as the error rate difference.

[0076] It should be noted that the preset stabilization time is set based on the signal locking time of the device under test receiving link and the minimum frame synchronization time required by the communication protocol, and the example value is 15 milliseconds.

[0077] S6.4: Determine whether the error rate difference meets the preset convergence determination parameter, and record the current received power value as the sensitivity test result when it meets the preset convergence determination parameter. It should be noted that the preset convergence determination parameter includes a convergence threshold and a minimum stable time length; the convergence threshold is set based on the allowed fluctuation range of the target error rate, and the example value is: when the target error rate is 10 -6 , take 10 -7 , and the target error rate statistics need to meet the law of large numbers; when the target error rate is 10 -6 , at least 10 -7 bits of transmission are required to obtain effective statistics; the minimum stable time length is set based on the minimum effective time window of the physical layer error statistics of the device under test, and the example value is: 100 milliseconds for Wi-Fi 6E devices, and the PPDU transmission and block confirmation mechanism of Wi-Fi 6E requires ≥100ms to complete a complete frame exchange.

[0078] Specifically, the error rate difference is compared with the convergence threshold in the preset convergence determination parameter, and whether the duration of the error rate difference being less than the convergence threshold exceeds the preset minimum stable time length is monitored; only when the error rate difference is less than the convergence threshold and the duration is greater than the minimum stable time length, the current output power value of the vector signal generator is read, and the current output power value is recorded as the sensitivity test result.

[0079] S6.5: Automatically generate a test report based on the sensitivity test result and the entire test process data.

[0080] Specifically, the sensitivity test result, the noise distribution parameter set, the key frame data of the high-precision dynamic noise map, the power adjustment time record, the bit error rate change, the to-be-tested device model and the test frequency band bandwidth parameter, the compensation transmission power calibration value time sequence, and the received power prediction value sequence are structured and organized according to a predefined report template, the sensitivity test result value is filled in a result section, the noise distribution parameter set and the high-precision dynamic noise map are embedded in a noise environment section, the power adjustment and the bit error rate change comparison graph is inserted in a power control section, and the device model and the frequency band bandwidth are listed in a device parameter section; and a test report is generated.

[0081] It should be noted that the structured organization of the predefined report template refers to classified filling of data according to a fixed section framework.

[0082] The embodiment also provides a computer device suitable for the sensitivity automatic test method of the wideband radio frequency system, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the sensitivity automatic test method of the wideband radio frequency system proposed in the above embodiment.

[0083] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to perform wired or wireless communication with an external terminal. The wireless communication can be achieved through WIFI, a carrier network, NFC (near field communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or can be a key, a trackball, or a touchpad arranged on the shell of the computer device. In addition, the input device can also be an external keyboard, a touchpad, a mouse, or the like.

[0084] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement the method for automatically testing sensitivity of a wideband radio frequency system as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0085] To sum up, the application can realize accurate perception and active compensation of the spatial noise of a test environment by generating a high-precision dynamic noise map and calculating a compensation transmit power calibration value. By fusing multi-source noise data to construct a visual electromagnetic environment model, the power pre-adjustment of a transmit signal has environment perception capability, so that the receive power of a device under test can be quickly approximated to the vicinity of a real sensitivity threshold in the initial stage of testing, thereby reducing the iteration number and time required for fine power adjustment. The test error caused by environmental noise fluctuation is effectively suppressed, and the accuracy and reliability of the sensitivity test result are improved, which lays a solid foundation for the acceleration and convergence of the entire automated test process and improves the test efficiency.

[0086] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application but not limit the application. Although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the application, and all of them should be covered in the scope of the claims of the application.

Claims

1. A method for automatic testing of sensitivity of a wideband radio frequency system, characterized by: The application relates to a method for testing sensitivity of a wireless communication device. The method comprises the following steps: starting a test procedure, coordinating multiple calibration devices and the wireless communication device in a wireless network to synchronously sample spectrum data of the calibration devices and process the spectrum data into a noise distribution parameter set in a same test time slot through an air interface; uploading the noise distribution parameter set to a master control end for fusion to generate a high-precision dynamic noise map; extracting a noise power value of a location where the wireless communication device is located from the high-precision dynamic noise map and obtaining a compensation transmission power calibration value; adjusting output power of a vector signal generator according to the compensation transmission power calibration value, generating a wideband test signal and transmitting the wideband test signal to the wireless communication device, and reading internal state information of a receiving link of the wireless communication device in real time; inputting the internal state information into a machine learning model to output a receiving power prediction value required to reach a target bit error rate under a current test condition; 2. The method of automatically testing the sensitivity of a wideband radio frequency system of claim 1, wherein: dynamically adjusting the output power of the vector signal generator according to the receiving power prediction value and synchronously monitoring a bit error rate change until a difference between the bit error rate and a preset target bit error rate threshold meets a convergence condition, recording a receiving power value at the convergence as a sensitivity test result and generating a test report. The method for synchronously sampling spectrum data of the calibration devices and processing the spectrum data into the noise distribution parameter set comprises the following steps: the master control end starts the test procedure, collects spatial position coordinates of the calibration devices, establishes nanosecond-level time synchronization with the calibration devices and the wireless communication device through a precise time protocol and allocates a same sampling window; the calibration devices perform radio frequency sampling in the same sampling window through a local oscillator frequency and a sampling rate to obtain sampling data; the sampling data is subjected to window processing to generate spectrum data of the calibration devices through fast Fourier transform; the spectrum data of the calibration devices is subjected to Gaussian distribution fitting to extract noise mean value parameters and noise variance parameters; 3. The method of automatically testing the sensitivity of a wideband radio frequency system of claim 1, wherein: the spatial position coordinates of the calibration devices, the noise mean value parameters and the noise variance parameters are combined to generate the noise distribution parameter set. The method for generating the high-precision dynamic noise map comprises the following steps: the master control end receives the noise distribution parameter set, performs standardization preprocessing to generate a standardized noise distribution parameter set; based on the standardized noise distribution parameter set, a Kriging spatial interpolation algorithm is adopted to calculate noise mean value estimation values and noise variance estimation values of any position in a test region to form initial noise distribution data; the initial noise distribution data is input into a time Kalman filter for time dimension filtering and optimization to generate optimized noise distribution data; 4. The method of automatically testing the sensitivity of a wideband radio frequency system of claim 1, wherein: the optimized noise distribution data is subjected to multi-resolution fusion processing to generate the high-precision dynamic noise map. The method for obtaining the compensation transmission power calibration value comprises the following steps: spatial position coordinates of the wireless communication device are collected, the high-precision dynamic noise map is queried and noise mean value estimation values and noise variance estimation values of a location where the wireless communication device is located are obtained; based on the noise mean value estimation values and the noise variance estimation values, a noise power synthesis algorithm is used to calculate a noise power value; taking the spatial position coordinates of the wireless communication device as a center, spatial data points and noise parameters in a preset neighborhood range are extracted from the high-precision dynamic noise map to form a local noise parameter set; According to the local noise parameter set, the spatial uncertainty weight factor is calculated, and the noise power value is corrected to obtain an initial compensation power calibration value; The initial compensation power calibration value is subjected to amplitude limiting and timing smoothing processing to obtain a compensation transmit power calibration value.

5. The method of automatically testing the sensitivity of a wideband radio frequency system of claim 1, wherein: The internal state information of the receiving link of the device under test is read in real time, and the specific steps are as follows, The compensation transmit power calibration value is converted through the SCPI protocol format to generate an instrument control instruction; The vector signal generator adjusts the output power according to the instrument control instruction, generates a wideband test signal through direct digital frequency synthesis and I / Q modulation algorithm; The device under test demodulates the wideband test signal through the internal receiving link, generates internal state information, and reads it in real time through the diagnostic interface.

6. The method of automatically testing the sensitivity of a wideband radio frequency system of claim 1, wherein: The internal state information includes automatic gain control voltage, channel estimation coefficient and equalizer tap weight.

7. The method of automatically testing the sensitivity of a wideband radio frequency system of claim 1, wherein: The output corresponds to the predicted value of the receiving power required to reach the target bit error rate under the current test condition, and the specific steps are as follows, The automatic gain control voltage, channel estimation coefficient and equalizer tap weight in the internal state information are combined with the historical state data to construct a spatio-temporal joint feature tensor; The spatio-temporal joint feature tensor is input into a machine learning model to extract a deep spatio-temporal feature vector; According to the deep spatio-temporal feature vector, the posterior probability distribution of the predicted value of the receiving power is calculated through Bayesian optimization reasoning, and the predicted value of the receiving power required to reach the target bit error rate is output.

8. The method of automatically testing the sensitivity of a wideband radio frequency system of claim 1, wherein: According to the predicted value of the receiving power, the output power of the vector signal generator is dynamically adjusted and the bit error rate is monitored synchronously until the difference between the bit error rate and the preset target bit error rate threshold meets the convergence condition, the receiving power value at the time of convergence is recorded as the sensitivity test result, and a test report is generated, and the specific steps are as follows, Based on the predicted value of the receiving power, the initial output power of the vector signal generator is set; Real-time monitoring of bit error rate data, according to the difference between the bit error rate data and the preset target bit error rate threshold, dynamically generating a power control strategy and adjusting the output power of the vector signal generator; According to the adjusted output power, the bit error rate change of the device under test is monitored synchronously, and the bit error rate difference between the new bit error rate data of the device under test and the preset target bit error rate threshold is calculated; Determine whether the bit error rate difference meets the preset convergence determination parameter, and record the current receiving power value as the sensitivity test result when it meets the requirement; According to the sensitivity test result and the entire test process data, a test report is automatically generated. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the method for automatically testing the sensitivity of the wideband radio frequency system according to any one of claims 1-8.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the method for automatically testing the sensitivity of the wideband radio frequency system according to any one of claims 1-8.

Images