Communication line noise filtering system and method

Abstract

The invention relates to the technical field of noise filtering, in particular to a communication line noise filtering system and method, and the system comprises a temperature coupling module, a voltage response module, a parameter joint debugging module and a link checking module. According to the method, inductance and capacitance node temperature data are collected and input into a piecewise linear regression model, the temperature-to-resonant frequency deviation gradient is calculated, a noise suppression bandwidth threshold value is dynamically matched in combination with a reference table, the frequency stability under temperature fluctuation is improved, Kalman filtering is used for processing voltage sampling, and a fluctuation predicted value is generated. According to the method, interval indexes are extracted based on a parameter mapping table, instant inductance and capacitance parameters are generated through least square fitting, voltage transient fluctuation self-adaption is achieved, noise power spectrum density is calculated through fast Fourier transform, a 3dB attenuation point is compared with a threshold value difference value to trigger parameter rollback, closed-loop control is formed, it is ensured that suppression bandwidth is continuously matched with a noise frequency band, and noise suppression is achieved. The bandwidth offset is reduced to be within 1.2%, and the optimization response time is shortened to 200 ms level.

Description

Technical Field

[0001] The present invention relates to the technical field of noise filtering, and in particular, to a communication line noise filtering system and method. Background Art

[0002] The technical field of noise filtering includes technical methods used in various communication and electronic systems to reduce or suppress unnecessary electrical noise interference. The core content of this technical field is to filter out interference signals in the communication line by selecting different types of filter circuits to ensure the integrity and stability of the communication signal. Noise filtering is usually based on filter structures formed by capacitors, inductors, and their combinations, and is widely used in wired communication, wireless communication, power transmission, and electronic devices. In terms of systematic applications, noise filtering technology covers various links such as source noise suppression, path noise isolation, and receiver noise cancellation, and selects appropriate filtering strategies according to the frequency characteristics and intensity of the noise. For example, circuit structures such as low-pass filters, high-pass filters, and band-pass filters are used to effectively process noise in different frequency bands.

[0003] Among them, a communication line noise filtering system refers to a hardware system or circuit structure applied to a wired communication path to reduce external electromagnetic interference or self-signal crosstalk. The patent theme mainly focuses on the noise problems caused by the complex electromagnetic interference environment in the communication line, covering methods such as using passive filters, active noise reduction circuits, power line filters, and signal line filters. Specifically, a low-pass filter combination, an LC filter, or a π-type filter is serially arranged between the communication line and the signal processing unit, and impedance matching technology is used to weaken the interference formed by the transmission of external electromagnetic waves on the communication line. At the same time, the interference conduction between lines is reduced by reasonably arranging the grounding device, so as to complete the suppression and management of the communication path noise.

[0004] In the prior art, the inductance and capacitance parameters of the fixed-structure filter cannot be automatically corrected with temperature changes. When the ambient temperature rises by 10 °C, the inductance value drift reaches 5%-8%, resulting in a shift of the resonant frequency and a suppression blind area of 3-5 MHz. The traditional impedance matching strategy does not consider the impact of transient voltage fluctuations. When the voltage mutation exceeds ±7%, the insertion loss curve of the filter shows non-linear distortion, resulting in a reduction of the effective suppression bandwidth of 2.4-3.6 dB. Manual parameter adjustment depends on offline spectrum test data. Each calibration requires interrupting the communication link and takes more than 30 minutes, and transient interference signals below 500 ms cannot be captured. The existing power spectrum detection uses a timed sampling mechanism, and the capture probability of intermittent pulse noise is less than 65%, with a missed detection rate of 15%-22% in burst interference scenarios. The grounding layout strategy is not linked with dynamic parameters, and the suppression efficiency of the interference conduction between lines decreases by 40%-55% as the temperature rises. Summary of the Invention

[0005] The object of the present invention is to solve the deficiencies existing in the prior art, and a communication line noise filtering system and method are proposed.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A communication line noise filtering system includes: A temperature coupling module, configured to collect the real-time temperature value of the inductor node and the real-time temperature value of the capacitor node through a dual-channel thermosensor, input the two into a piecewise linear regression model to calculate the deviation gradient of the temperature with respect to the resonant frequency, generate a temperature deviation gradient value, and transfer the temperature deviation gradient value and a preset resonant frequency reference table to a voltage response module; A voltage response module, configured to receive the temperature deviation gradient value and the resonant frequency reference table, match the noise suppression bandwidth threshold corresponding to the current temperature based on the resonant frequency reference table, detect the input voltage amplitude volatility, call a Kalman filtering algorithm to process the voltage sampling sequence to generate a voltage fluctuation prediction value, and transfer it to a parameter joint adjustment module; A parameter joint adjustment module, configured to receive the voltage fluctuation prediction value, extract an interval index from a preset inductor-capacitor parameter mapping table according to the voltage fluctuation prediction value, perform linear fitting of the least squares method on the inductor value to generate an instantaneous inductor parameter, perform the same operation on the capacitor value to generate an instantaneous capacitor parameter, and transfer the instantaneous inductor parameter and the instantaneous capacitor parameter to a link verification module.

[0007] As a further solution of the present invention, the temperature deviation gradient value includes a temperature measurement error correction factor, a resonant frequency change compensation coefficient, and a node thermal response characteristic parameter, the voltage fluctuation prediction value includes a voltage stability index, a voltage fluctuation trend parameter, and a voltage noise prediction value, the instantaneous inductor parameter includes a dynamic inductor adjustment coefficient, a real-time temperature drift calibration amount, and an inductor tolerance correction amount, and the instantaneous capacitor parameter includes a dynamic capacitor adjustment coefficient, a real-time frequency drift calibration amount, and a capacitor tolerance correction amount.

[0008] As a further solution of the present invention, the piecewise linear regression model dynamically divides piecewise nodes based on the gradient change rate of the inductor and capacitor temperature intervals, and sets weight coefficients for linear regression within multiple intervals; The Kalman filtering algorithm constructs a state equation through an ambient temperature interference correction factor, a voltage attenuation dynamic coefficient, and a voltage sequence dispersion index; The inductor-capacitor parameter mapping table divides the upper and lower limits of the fluctuation interval based on 1.2 - 1.5 times the average value of the voltage volatility.

[0009] As a further solution of the present invention, the temperature coupling module includes: The temperature acquisition sub-module detects the temperature sensor signals of the inductor node and the capacitor node, adopts a dual-channel signal synchronous acquisition circuit, sets the signal sampling interval and the noise filtering threshold, calibrates the amplitude and aligns the timing of the original signal, records the inductor temperature value and the capacitor temperature value respectively, establishes two groups of real-time temperature time series data with synchronized timestamps, and generates dual-channel temperature values; The regression processing sub-module divides the segmentation nodes of the inductor and capacitor temperature ranges based on the dual-channel temperature values, sets the weight coefficients of linear regression in multiple ranges, takes the inductor temperature value as the independent variable and the capacitor temperature value as the covariate and inputs them into the model, calculates the slope change amount of the temperature and the resonance frequency in multiple segmented ranges, extracts the set of slope parameters of the segmented ranges, and generates a set of regression coefficients; The segmentation nodes are dynamically adjusted when the gradient change rate of adjacent temperature sampling points exceeds 0.5 °C / s; The gradient calculation sub-module calls the set of regression coefficients, combines the segmented range to which the current dual-channel temperature value belongs, selects the corresponding slope parameter, multiplies the difference between the dual-channel temperature value and the reference frequency value by the slope parameter according to the temperature-frequency mapping relationship in the resonance frequency reference table, outputs the frequency offset caused by unit temperature change, and generates a temperature deviation gradient value.

[0010] As a further solution of the present invention, the voltage response module includes: The threshold matching sub-module, based on the temperature deviation gradient value, calls the temperature-frequency mapping relationship in the resonance frequency reference table, performs cubic spline interpolation calculation on the current temperature value, matches the nearest neighbor reference frequency point, extracts the upper and lower calibration values of the noise suppression bandwidth corresponding to the frequency point, and generates a bandwidth threshold matching value; The cubic spline interpolation satisfies that the first derivative is continuous at the nodes and the second derivative error is less than 0.1 Hz / °C; The fluctuation detection sub-module detects the peak-valley difference of the amplitude of the input voltage signal, intercepts the sampling sequence with a fixed time window, calculates the cumulative sum of the absolute values of the amplitude differences between adjacent sampling points in the window, and divides the cumulative sum by the product of the time window length and the reference amplitude to generate a voltage fluctuation rate; The filtering prediction sub-module calls the bandwidth threshold matching value and the voltage fluctuation rate, and uses the formula: ; Performs an operation to obtain the voltage fluctuation prediction component, superimposes it on the prediction covariance matrix in the Kalman filter state equation, and generates a voltage fluctuation prediction value; Among them, represents the voltage fluctuation prediction component, represents the temperature deviation gradient value, represents the resonance frequency reference value, represents the noise suppression bandwidth threshold, represents the voltage fluctuation rate, It represents the environmental temperature interference correction factor, which is obtained by calibrating the contribution degree of temperature interference to voltage fluctuation through experiments. It represents the voltage attenuation dynamic coefficient, which is obtained by regression analysis and fitting of historical voltage data. It represents the voltage sequence dispersion index, which is the ratio of the standard deviation to the mean of the voltage sampling sequence.

[0011] As a further solution of the present invention, the parameter joint debugging module includes: The fluctuation interval indexing sub-module collects the predicted voltage fluctuation value. According to the interval division rule defined in the preset inductance-capacitance parameter mapping table, it compares the predicted value with the upper and lower limit values of the fluctuation interval in the mapping table item by item, determines the index code of the interval to which the predicted value belongs, and generates a fluctuation interval index. The inductance parameter generation sub-module calls the fluctuation interval index, extracts the set of inductance values corresponding to the index in the mapping table, constructs a linear equation system with the index sequence as the independent variable and the inductance value as the dependent variable, calculates the minimum value of the sum of squared residuals of the equation system by the least squares method, and solves the slope and intercept parameters of the inductance value to generate the instantaneous inductance parameter. The iteration termination condition of the least squares method is that the sum of squared residuals is less than 0.01 or the number of iterations exceeds 50 times. The capacitance parameter generation sub-module calls the fluctuation interval index, extracts the set of capacitance values corresponding to the index in the mapping table, and based on the linear relationship between the index sequence and the capacitance value, applies the same least squares method iteration process as the inductance parameter generation sub-module to calculate the slope and intercept parameters of the capacitance value and generate the instantaneous capacitance parameter.

[0012] As a further solution of the present invention, the system further includes: The link verification module is used to receive the instantaneous inductance parameter and the instantaneous capacitance parameter, calculate the noise power spectral density at the output end of the filter through fast Fourier transform, compare the difference between the 3dB attenuation point of the power spectral density and the noise suppression bandwidth threshold. If the difference exceeds the threshold, it triggers a parameter rollback instruction and feedbacks it to the parameter joint debugging module to regenerate the parameters, outputs the effective inductance parameter and the effective capacitance parameter, and loads them to the filter control unit.

[0013] As a further solution of the present invention, the effective inductance parameter and the effective capacitance parameter include the noise suppression accuracy index, the filter performance verification result, and the link signal integrity evaluation value. The difference threshold is calibrated to be ±10% of the noise suppression bandwidth threshold through spectral analysis experiments.

[0014] As a further solution of the present invention, the link verification module includes: The noise spectrum calculation sub-module calls the instantaneous inductance parameter and the instantaneous capacitance parameter, obtains the noise time-domain signal sequence at the output end of the filter, applies the fast Fourier transform to convert the time-domain signal into frequency-domain components, calculates the ratio of the square of the amplitude of the frequency-domain components to the frequency bandwidth, and generates the noise power spectral density; The threshold verification sub-module, based on the noise power spectral density, locates the frequency point where the amplitude in the power spectrum decays to 70.7% of the maximum amplitude as the 3dB attenuation point, extracts the preset noise suppression bandwidth threshold, calculates the absolute difference between the attenuation point and the threshold, compares the difference with the threshold allowable range, determines whether the difference exceeds the upper limit of the range, and generates a difference verification result; The threshold allowable range is calibrated as [85%, 115%] through the spectral analysis experiment of 500 groups of noise samples; The parameter effective sub-module calls the difference verification result, executes branch logic according to the over-limit state. If it is over-limit, it clears the current parameters and triggers a rollback instruction. If it is not over-limit, it marks the parameter effective state and generates effective inductance parameters and effective capacitance parameters.

[0015] A communication line noise filtering method, which is executed based on the above communication line noise filtering system, includes the following steps: S1: Respectively collect the real-time temperature values of the inductance node and the capacitance node through a dual-channel temperature sensor, input the two groups of temperature values into a piecewise linear regression model to calculate the temperature deviation gradient of the resonance frequency, and generate a temperature deviation gradient value; S2: Query the preset resonance frequency reference table based on the temperature deviation gradient value, match the noise suppression bandwidth threshold corresponding to the current temperature, detect the input voltage amplitude volatility at the same time, and call the Kalman filtering algorithm to perform prediction processing on the voltage sampling sequence to generate a voltage fluctuation prediction value; S3: Locate the interval index in the inductance-capacitance parameter mapping table according to the voltage fluctuation prediction value, perform a linear fitting operation of the least squares method on the inductance value to generate an instantaneous inductance parameter, and synchronously perform the same operation on the capacitance value to generate an instantaneous capacitance parameter; S4: For the instantaneous inductance parameter and the instantaneous capacitance parameter, calculate the noise power spectral density at the output end of the filter through the fast Fourier transform, compare the difference between the 3dB attenuation point of the power spectral density and the noise suppression bandwidth threshold. If the difference exceeds the preset range, trigger a parameter rollback instruction to regenerate the parameters. Otherwise, output the effective inductance parameters and effective capacitance parameters to the filter control unit.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by collecting the temperature data of the inductor and capacitor nodes and inputting them into a piecewise linear regression model, calculating the deviation gradient of temperature with respect to the resonant frequency, and dynamically matching the noise suppression bandwidth threshold in combination with the resonant frequency reference table, the frequency stability of the filter in a temperature fluctuation scenario is improved. The Kalman filtering algorithm is used to process the voltage sampling sequence to generate a fluctuation prediction value, the interval index is extracted based on the inductor-capacitor parameter mapping table, and the instantaneous inductor and capacitor parameters are generated by least squares fitting to achieve parameter self-adaptation during voltage transient fluctuations. The fast Fourier transform is used to calculate the noise power spectral density at the output end, and the difference between the 3dB attenuation point and the threshold is compared to trigger parameter rollback, forming a closed-loop control structure to ensure continuous matching of the suppression bandwidth and the noise frequency band. This solution couples the analysis of temperature gradient, voltage fluctuation, and frequency domain characteristics, reducing the offset of the noise suppression bandwidth in a dynamic environment to less than 1.2% and shortening the parameter optimization response time to the 200ms level. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is the system flow chart of the present invention; Figure 2 is the flow chart of the temperature coupling module of the present invention; Figure 3 is the flow chart of the voltage response module of the present invention; Figure 4 is the flow chart of the parameter joint debugging module of the present invention; Figure 5 is the flow chart of the link verification module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the present invention. In addition, in the description of the present invention, "a plurality of" means two or more, unless otherwise specifically defined.

[0020] Embodiment 1

[0021] Please refer to Figure 1 , the present invention provides a technical solution: A communication line noise filtering system includes: A temperature coupling module, which is used to collect the real-time temperature value of the inductor node and the real-time temperature value of the capacitor node through a dual-channel thermal sensor, input both into a piecewise linear regression model to calculate the deviation gradient of temperature with respect to the resonant frequency, generate a temperature deviation gradient value, and transfer the temperature deviation gradient value and a preset resonant frequency reference table to a voltage response module; A voltage response module, which is used to receive the temperature deviation gradient value and the resonant frequency reference table, match the noise suppression bandwidth threshold corresponding to the current temperature based on the resonant frequency reference table, detect the input voltage amplitude volatility, call the Kalman filtering algorithm to process the voltage sampling sequence to generate a voltage fluctuation prediction value, and transfer it to a parameter joint adjustment module; A parameter joint adjustment module, which is used to receive the voltage fluctuation prediction value, extract an interval index from a preset inductor-capacitor parameter mapping table according to the voltage fluctuation prediction value, perform a linear fitting of the least squares method on the inductor value to generate an instant inductor parameter, perform the same operation on the capacitor value to generate an instant capacitor parameter, and transfer the instant inductor parameter and the instant capacitor parameter to a link verification module; A link verification module, which is used to receive the instant inductor parameter and the instant capacitor parameter, calculate the noise power spectral density at the output end of the filter through fast Fourier transform, compare the difference between the 3dB attenuation point of the power spectral density and the noise suppression bandwidth threshold, and if the difference exceeds the threshold, trigger a parameter rollback instruction and feedback it to the parameter joint adjustment module to regenerate the parameters, and output an effective inductor parameter and an effective capacitor parameter and load them into the filter control unit.

[0022] The temperature deviation gradient value includes a temperature measurement error correction factor, a resonant frequency change compensation coefficient, and a node thermal response characteristic parameter. The voltage fluctuation prediction value includes a voltage stability index, a voltage fluctuation trend parameter, and a voltage noise prediction value. The instant inductor parameter includes a dynamic inductor adjustment coefficient, a real-time temperature drift calibration amount, and an inductor tolerance correction amount. The instant capacitor parameter includes a dynamic capacitor adjustment coefficient, a real-time frequency drift calibration amount, and a capacitor tolerance correction amount. The effective inductor parameter and the effective capacitor parameter include a noise suppression accuracy index, a filter performance verification result, and a link signal integrity evaluation value.

[0023] The piecewise linear regression model dynamically divides piecewise nodes based on the gradient change rate of the inductor and capacitor temperature intervals, and sets the weight coefficients of the linear regression within multiple intervals; The Kalman filtering algorithm constructs a state equation through an environmental temperature interference correction factor, a voltage attenuation dynamic coefficient, and a voltage sequence dispersion index; The inductor-capacitor parameter mapping table divides the upper and lower limits of the fluctuation interval based on 1.2 - 1.5 times the average value of the voltage volatility; The difference threshold is calibrated through a spectrum analysis experiment to be ±10% of the noise suppression bandwidth threshold.

[0024] Please refer to Figure 2, the temperature coupling module includes: The temperature acquisition sub-module detects the temperature sensor signals of the inductor node and the capacitor node, adopts a dual-channel signal synchronous acquisition circuit, sets the signal sampling interval and the noise filtering threshold, calibrates the amplitude and aligns the time sequence of the original signal, records the inductor temperature value and the capacitor temperature value respectively, establishes two groups of real-time temperature time sequences with synchronized timestamps, and generates dual-channel temperature values. The function of the temperature acquisition sub-module is to accurately and synchronously obtain the operating temperatures of the inductor and the capacitor. First, high-precision temperature sensors, such as PT1000 platinum resistance thermometers, are deployed on the critical heat dissipation paths adjacent to the inductor and the capacitor. These sensors continuously monitor the temperature of the target nodes and output corresponding resistance or voltage signals. The dual-channel signal synchronous acquisition circuit is triggered, and this circuit uses a shared high-frequency clock signal (set to 100 kHz) to ensure that the sampling moments of the two sensor signal channels are strictly consistent, avoiding errors introduced by sampling time differences. The signal sampling interval is set to 10 milliseconds, that is, the system reads each sensor 100 times per second.

[0025] The acquired original signal is first subjected to noise filtering. The noise filtering threshold is set to 0.1 °C, which is determined based on the analysis of the inherent noise characteristics of the sensor and the normal operating temperature change rate, aiming to filter out fast and random fluctuations with an amplitude lower than 0.1 °C generated by electromagnetic interference or the sensor itself. Specifically, when executing, the reading of the current sampling point is compared with the reading of the previous sampling point . If the absolute difference between the two is less than 0.1 °C, that is , it is determined that is noise, and its value is corrected to . This process is independently applied to the temperature signals of the inductor and the capacitor.

[0026] After completing the noise filtering, the signal enters the amplitude calibration stage. Using the sensor calibration data table stored in the non-volatile memory (this table is generated based on factory calibration and periodic on-site multi-point temperature calibration, and contains the mapping relationship between the accurate output values of the sensor at standard temperatures of 0 °C, 25 °C, 50 °C, 75 °C, and 100 °C and the corresponding temperatures), the filtered reading is processed. Through the piecewise linear interpolation algorithm, according to the specific interval where the reading falls in the calibration data table, the accurate Celsius temperature value is calculated. Taking the inductor temperature as an example, if its filtered reading is 1.185 V, and it is found from the calibration table that the output is 1.150 V at 50 °C and 1.200 V at 60 °C, then the calibrated temperature is calculated as . The same calibration process is performed on the capacitor temperature .

[0027] Finally, perform timing alignment confirmation. Although the hardware design ensures synchronous acquisition, the software layer will double-check the timestamps of the two channels of data. . Taking the timestamp of the inductance temperature channel as the reference, fine-tune (usually within the microsecond level) the timestamp of the capacitance temperature data points to ensure that the and in the database strictly correspond to the same physical moment. The calibrated and aligned inductance temperature values and capacitance temperature values , together with the accurate timestamp , are structuredly recorded to form a continuous time series data stream, such as [(t1, 57.0, 54.2), (t2, 57.1, 54.3), (t3, 57.1, 54.4),...]. This sub-module finally outputs the dual-channel temperature values at the current moment, that is, the inductance temperature and capacitance temperature .

[0028] The regression processing sub-module divides the segmentation nodes of the inductance and capacitance temperature intervals based on the dual-channel temperature values, sets the weight coefficients of the linear regression within multiple intervals, takes the inductance temperature value as the independent variable and the capacitance temperature value as the covariate and inputs them into the model, calculates the slope change of the temperature and resonance frequency within multiple segmented intervals, extracts the set of slope parameters of the segmented intervals, and generates a set of regression coefficients; The segmentation nodes are dynamically adjusted when the gradient change rate of adjacent temperature sampling points exceeds 0.5 °C / s; The regression processing sub-module receives the dual-channel temperature time series data continuously output by the temperature acquisition sub-module. Its core task is to establish a mathematical model of the dynamic relationship between temperature and resonance frequency. First, it is necessary to determine the interval breakpoints of the piecewise linear model that describes this relationship, that is, the segmentation nodes. The initial nodes are set according to the device specification and historical operation data analysis, covering the expected operating temperature range, and are set as [20 °C, 30 °C, 40 °C, 50 °C, 60 °C, 70 °C, 80 °C]. These nodes are not fixed but are dynamically adjusted according to real-time data.

[0029] The basis for dynamic adjustment is the rate of temperature change. Calculate the gradient change rate of the temperature within the last N sampling points (set N = 5, corresponding to a 50 ms time window). For the inductance temperature sequence and capacitance temperature sequence respectively calculate the temperature change rate ) of adjacent sampling points (time interval and 。Further calculate the changes of these change rates themselves, that is, the second derivative or the difference of the change rates. Set a dynamic adjustment threshold: when the average gradient change rate of any one-way temperature exceeds 0.5 °C / s at M consecutive points (set M = 3), the system determines that the temperature change enters the fast dynamic period. The gradient change rate threshold of 0.5 °C / s is set based on the statistical analysis of the system's thermal inertia and the temperature rise rate under normal working modes. Exceeding this value indicates that it may enter the non-linear response area or a rapid environmental change. At this time, the system will insert a new segmentation node near the current temperature point or adjust the position of the existing node to more precisely capture the characteristics of the rapidly changing area. For example, if the inductor temperature rises from 48.8 °C to 49.0 °C in 30 ms (3 sampling points), the average gradient is (49.0 - 48.8) / 0.03 = 6.67 °C / s, far exceeding 0.5 °C / s, then the node is adjusted or added near 49 °C. Suppose after adjustment, the current valid nodes are [20, 30, 40, 49, 58, 70, 80] °C.

[0030] Next, assign the weight coefficients of linear regression to each temperature interval determined by the nodes 。The setting of the weight coefficients reflects the confidence level of model fitting or data quality in different temperature intervals. It is set with reference to the analysis results of the distribution density and linear fitting goodness ( value) of the (temperature, frequency) data points in each temperature interval in the historical database. The historical data points in a certain interval are dense and the value is close to 1, indicating that the relationship in this interval is stable and a higher weight is assigned; on the contrary, a lower weight is assigned to the interval with sparse data or a lower value. The specific setting process: Analyze the historical data, calculate the value of the interval [40 °C, 49 °C) as 0.98, and set it as the reference weight , the value of the interval [49 °C, 58 °C) is 0.92, then its weight is set as 。Set the weights of all intervals

[0031] Then, take the inductor temperature value collected in real time as the independent variable, the capacitor temperature value as the covariate, and the resonant frequency value measured at the corresponding moment through other modules of the system, and jointly input them into the piecewise weighted linear regression model. For the th temperature interval (defined by dynamic nodes, such as [49 °C, 58 °C)), the model is expressed as 。Utilize all valid historical and real-time data points {( )} within this interval, and combine with the weight coefficient , the weighted least squares method is used to solve the model parameters. The goal of the weighted least squares method is to minimize the weighted sum of squared residuals . By solving this optimization problem, the slopes of the resonant frequency with respect to the inductor temperature and the capacitor temperature within this interval are obtained and . This calculation is repeated for all N segmented intervals to obtain a complete set of slope parameters {( ),( ),…,( )}. Suppose for the interval [49°C, 58°C), the calculated slopes are and . Finally, this sub-module generates a set of regression coefficients containing the slope parameters for all segmented intervals

[0032] The gradient calculation sub-module calls the set of regression coefficients, combines the segmented interval to which the current dual-channel temperature value belongs, selects the corresponding slope parameters, multiplies the difference between the dual-channel temperature value and the reference frequency value by the slope parameters according to the temperature-frequency mapping relationship in the resonant frequency reference table, outputs the frequency offset caused by a unit temperature change, and generates a temperature deviation gradient value

[0033] When the gradient calculation sub-module starts, it first retrieves the set of regression coefficients {( ),…,( )} updated in real-time by the regression processing sub-module. At the same time, it obtains the current latest dual-channel temperature values, namely and from the temperature acquisition sub-module. Let the current inductor temperature be , and the capacitor temperature be

[0034] The system determines the segmented interval to which the current temperature value belongs based on the dynamically adjusted segmented node list [20, 30, 40, 49, 58, 70, 80]°C. Since and , both temperature values fall within the 3rd effective interval [49°C, 58°C) (indexed from 0, so it is the 3rd interval, or the 4th interval in the list order). Therefore, the system accurately selects the pair of slope parameters corresponding to this interval from the set of regression coefficients ([[]] ). According to the calculation results of the previous sub-module, this pair of parameters is and

[0035] Next, calculations need to be combined with the resonant frequency reference at the current temperature. The system consults the internally stored resonant frequency reference table, which is calibrated by a precision network analyzer in a high-precision temperature control box and records the ideal resonant frequencies at specific reference temperature points

[0036] ​​Table 1: Example of the Temperature-Frequency Mapping Relationship Reference Table

[0037] As shown in Table 1, this table lists the reference values of the resonant frequencies corresponding to some discrete reference temperature points.

[0038] The core task of this sub-module is to calculate a gradient value that represents the frequency shift caused by a unit temperature change in the current state. This gradient value reflects the overall sensitivity of the system frequency to temperature changes. The calculation method is based on the selected slope parameter ( ). Considering that the inductor is usually the main component affecting frequency stability, and the absolute value of its temperature coefficient is greater than that of the capacitor , in this embodiment, the absolute value of the slope of the inductor temperature is selected as the dominant gradient. Therefore, the frequency shift caused by a unit temperature change in the output is determined to be . This value is named the temperature deviation gradient value . Finally, this sub-module generates the temperature deviation gradient value for use by subsequent modules.

[0039] Please refer to Figure 3 , the voltage response module includes: Based on the temperature deviation gradient value, the threshold matching sub-module calls the temperature-frequency mapping relationship in the resonant frequency reference table, performs cubic spline interpolation calculation on the current temperature value, matches the nearest neighbor reference frequency point, extracts the upper and lower calibration values of the noise suppression bandwidth corresponding to the frequency point, and generates a bandwidth threshold matching value; The cubic spline interpolation satisfies that the first derivative is continuous at the nodes and the second derivative error is less than 0.1 Hz / °C; The threshold matching sub-module receives the temperature deviation gradient value output by the gradient calculation sub-module . It combines the current real-time temperature and calls the stored resonant frequency reference table (Table 1) for calculation. First, calculate the average temperature of the current working state .

[0040] According to the discrete (temperature, frequency) data points in the reference table (Table 1), for Precisely interpolate the reference resonant frequency at the specified point. Use the cubic spline interpolation method, which can ensure that the interpolation curve is not only continuous at all original data nodes but also has a continuous first derivative (slope). At the same time, the selected cubic spline interpolation implementation needs to meet an additional smoothness constraint: for the interpolation polynomial segment between any two data nodes, the rate of change of its second derivative (representing the degree of curve bending) with respect to temperature must be less than 0.1 Hz / °C². This constraint is based on the understanding of the generally smooth variation characteristics of the frequency-temperature curve of the physical system. Setting an upper limit of 0.1 Hz / °C² is to suppress the sharp bends that may be introduced by interpolation and do not conform to physical laws. Using the data points around in the reference table (40°C, 13.5615 MHz), (50°C, 13.5600 MHz), (60°C, 13.5580 MHz), (70°C, 13.5570 MHz), perform cubic spline interpolation through a standard numerical calculation library to obtain the corresponding interpolated reference frequency . The calculation result is .

[0041] After obtaining the accurate current reference frequency, match it with the discrete frequency values in the reference table (Table 1) to find the reference point that is numerically closest. Calculate the absolute value of the difference between it and each frequency in the table: the difference from 13.5600 MHz (50°C) is , and the difference from 13.5580 MHz (60°C) is . By comparison, 13.5600 MHz is the nearest neighbor reference frequency point, .

[0042] Finally, the system accesses an internal "frequency-bandwidth" calibration database. This database stores the calibration values of the filter bandwidth required to achieve the target noise suppression performance for different reference center frequency points. These calibration values are determined through offline experiments: for multiple filter prototypes with the center frequency tuned to , apply a standard noise signal, repeatedly adjust its parameters (i.e., adjust the bandwidth) and measure the output noise power until the noise suppression index (such as out-of-band rejection ) meets the standard, and record the corresponding 3 dB bandwidth at this time. Looking up the database, for , the corresponding calibrated noise suppression bandwidth is . This sub-module finally generates a bandwidth threshold matching value , which will be used as the target bandwidth reference for subsequent calculations.

[0043] The fluctuation detection sub-module detects the peak-valley difference of the amplitude of the input voltage signal, intercepts the sampling sequence with a fixed time window, calculates the sum of the absolute values of the amplitude differences between adjacent sampling points within the window, and divides the sum by the product of the time window length and the reference amplitude to generate the voltage fluctuation rate. The fluctuation detection sub-module continuously monitors the voltage signal input to the system . It intercepts and analyzes the signal with a fixed time window to quantify its degree of fluctuation. Set the time window length . The selection of this window length is based on a balanced consideration of the system's dynamic response time and the frequency characteristics of typical interference signals. 100 ms can capture most power frequency interferences and harmonics, and at the same time, it is not too long for the response of the control loop. The sampling rate of the voltage signal is set to , which means that within a 100 ms window, a total of discrete voltage sample points are collected.

[0044] The obtained voltage sampling sequence is denoted as . A specific sampling sequence segment can be: 3.05V, 3.08V, 3.02V, 3.06V, 3.10V, 3.07V, 3.03V, …, 3.09V. The sub-module then calculates the sum of the absolute values of the amplitude differences between all adjacent sampling points within this window. The calculation process is . Calculate the first few differences of the above sequence segment: . After traversing all 99 differences within the window and accumulating them, the total sum is obtained.

[0045] At the same time, calculate the arithmetic mean of all 100 voltage samples within this time window as the reference amplitude during this window period . The calculation formula is . After summing all the samples within the window and dividing by 100, is obtained.

[0046] Finally, calculate the voltage fluctuation rate according to the formula . This formula is defined as the sum of the absolute differences divided by the product of the time window length (in seconds) and the reference amplitude: . Substitute the values: . This value has the unit of , which reflects the cumulative relative amplitude of voltage jumps per unit time. The sub-module finally generates the voltage fluctuation rate .

[0047] The filtering prediction sub-module calls the bandwidth threshold matching value and the voltage fluctuation rate, and uses the formula: ; The operation obtains the voltage fluctuation prediction component, superimposes the predicted covariance matrix in the Kalman filter state equation, and generates the voltage fluctuation prediction value; Among them, represents the voltage fluctuation prediction component, represents the temperature deviation gradient value, represents the resonant frequency reference value, represents the noise suppression bandwidth threshold, represents the voltage volatility, represents the environmental temperature interference correction factor, which is obtained by calibrating the contribution degree of temperature interference to voltage fluctuation through experiments, represents the voltage attenuation dynamic coefficient, which is obtained by regression analysis and fitting of historical voltage data, represents the voltage sequence dispersion index, which is the ratio of the standard deviation to the mean of the voltage sampling sequence.

[0048] The filtering prediction sub-module integrates the information from the previous steps and calculates a comprehensive voltage fluctuation prediction component , which is designed to quantify the overall disturbance level faced by the system and is used to guide the subsequent adjustment of the Kalman filter. It calls the bandwidth threshold matching value and the voltage volatility , and applies the following formula: , this formula combines the effects of temperature drift and inherent voltage fluctuation. The parameters in the formula are explained and valued as follows: : The temperature deviation gradient value, provided by the gradient calculation sub-module, reflects the sensitivity of frequency to temperature, .

[0049] : The current average temperature, calculated by the threshold matching sub-module, .

[0050] : The nominal or design reference operating temperature of the system, which is a fixed design parameter, set to .

[0051] : The noise suppression bandwidth threshold, provided by the threshold matching sub-module, which is the target bandwidth of the filter, .

[0052] : The voltage attenuation dynamic coefficient, which is a dimensionless parameter, obtained by fitting the time series data of the historical voltage volatility using the autoregressive model (AR(1)), and reflects the coefficient of fluctuation persistence. The specific operation is: collect the calculated per second within the past few hoursValue sequence, fitting , and the obtained coefficient is what is required. Through this process, is obtained.

[0053] : Relative voltage volatility, which is a dimensionless index and is used to represent voltage fluctuation in the formula. It is calculated from the data of the fluctuation detection sub-module: .

[0054] : Ambient temperature interference correction factor, dimensionless, quantifying the influence degree of ambient temperature on voltage relative fluctuation. It is calibrated through controlled experiments: at different stable ambient temperatures (from 20°C to 70°C, with a step of 10°C), run the system for a long time, and record the average value of the voltage sequence dispersion index at each temperature. Plot the scatter diagram of ( ), and perform linear fitting . The obtained slope is what is sought. The experimental calibration result is .

[0055] : Voltage sequence dispersion index, dimensionless, defined as the standard deviation of the voltage sampling within the current time window divided by its mean value (i.e., the coefficient of variation CV). For the voltage window data in Paragraph 5, calculate its standard deviation as , and the mean value as , then .

[0056] Now, substitute these values into the formula for calculation. Pay attention to handling unit consistency: The first term (temperature contribution ): Calculate the absolute frequency shift caused by temperature . Bandwidth . For ratio calculation, the units need to be unified. Conversion rule: Convert from kHz to Hz by multiplying by the conversion factor 1000 Hz / kHz. . Then . This is a dimensionless ratio.

[0057] The second term (voltage contribution ): All parameters are dimensionless. Calculate . This is also a dimensionless value.

[0058] The total predicted component of voltage fluctuation .

[0059] The calculated is a comprehensive dimensionless perturbation index. This value is then used to adjust the prediction covariance matrix of the Kalman filter . Specifically, the process noise covariance is updated to make it related to , for example , where is the basic process noise setting, and is a regulation factor. When the value increases, the process noise increases accordingly, making the Kalman filter trust the current measurement value more, more adaptable, and able to track the changes in the system state faster. After this adjustment, the Kalman filter outputs an optimized estimate and prediction of the system state (such as voltage, frequency, etc.), and this output is the voltage fluctuation prediction value

[0060] This calculation result is relatively small (refer to the interval definition in Table 2 below), indicating that the overall perturbation level of the current system is low. This value itself is not the final output voltage prediction, but is passed as a key input to the subsequent parameter joint adjustment module for selecting the appropriate L / C parameter range

[0061] Please refer to Figure 4 , the parameter joint adjustment module includes: The fluctuation interval index sub-module collects the voltage fluctuation prediction value, and according to the interval division rules defined in the preset inductance-capacitance parameter mapping table, compares the prediction value item by item with the upper and lower limit values of the fluctuation interval of the mapping table to determine the index code of the interval to which the prediction value belongs, and generates a fluctuation interval index The fluctuation interval index sub-module receives the voltage fluctuation prediction value calculated by the filtering prediction sub-module . Its task is to map this continuous value to a discrete interval index, which corresponds to the preset inductance-capacitance parameter adjustment strategy. This mapping is performed according to the fluctuation interval division rules defined in the inductance-capacitance parameter mapping table stored in the system (shown in Table 2 below)

[0062] Table 2: Example of inductance-capacitance parameter mapping table

[0063] As shown in Table 2, this table is determined based on system design and simulation experiments, and defines different fluctuation prediction value intervals, corresponding fluctuation levels, and recommended inductance and capacitance parameter value ranges

[0064] The system will use the current value Compare the boundary values ​​of each interval in Table 2 in sequence. First, determine .because , this condition is met. Therefore, the current voltage fluctuation prediction value It is determined to belong to the "low" volatility range with index code 1. The submodule then generates the result: the volatility range index is 1.

[0065] The inductance parameter generation submodule calls the fluctuation interval index, extracts the inductance value set corresponding to the index in the mapping table, constructs a linear equation group with the index sequence as the independent variable and the inductance value as the dependent variable, calculates the minimum value of the residual square sum of the equation group through the least squares method, solves the slope and intercept parameters of the inductance value, and generates the instant inductance parameters; The least squares method iteration termination condition is that the residual sum of squares is less than 0.01 or the number of iterations exceeds 50; The inductor parameter generation submodule receives the fluctuation interval index 1. Based on this index, it first extracts the corresponding recommended inductor value set from the inductor-capacitor parameter mapping table (Table 2), that is, [95nH, 105nH]. In order to determine an accurate instantaneous inductor value within this range, the submodule uses the stored historical operation data. This data is a paired sequence of [(index, actual effective inductor value / nH)], which records the inductor values ​​used by the system when it was working stably under different fluctuation intervals over a period of time in the past.

[0066] Assume that the most recent 7 sets of historical data are: X (index) = [2, 1, 1, 3, 2, 1, 4], Y (inductance value / nH) = [108, 102, 100, 118, 111, 103, 128]. The system builds a linear model based on these historical data. , trying to find the linear trend between the index and the optimal inductance value. The least squares method is used to calculate the model parameters (slope) and (intercept). The principle of least squares method is to find the parameter So that the predicted value of the historical data point ( ) and the actual value The sum of squared errors between Minimum. Solved by standard analytical solution or numerical optimization method, we get nH / index unit and nH. The solution process sets the iteration termination condition: it stops when the change of the residual sum of squares is less than 0.01nH² or the number of iterations exceeds 50. These conditions are set based on the trade-off between convergence speed and computing resources.

[0067] After obtaining the model parameters, use the current fluctuation range index 1 to substitute into the model for prediction: . Check the predicted values Whether it falls within the recommended range [95 nH, 105 nH] corresponding to Index 1. The result is within the range. Therefore, the system adopts this predicted value as the instantaneous inductance parameter. This sub-module finally generates the instantaneous inductance parameter .

[0068] The capacitance parameter generation sub-module calls the fluctuation range index, extracts the set of capacitance values corresponding to the index in the mapping table, and based on the linear relationship between the index sequence and the capacitance values, applies the same least squares iterative process as the inductance parameter generation sub-module to calculate the slope and intercept parameters of the capacitance values, and generates the instantaneous capacitance parameter.

[0069] The capacitance parameter generation sub-module also receives the fluctuation range index 1. It extracts the set of recommended capacitance values corresponding to Index 1 from the inductance-capacitance parameter mapping table (Table 2), that is, [130 pF, 140 pF]. Similar to the inductance parameter generation, this sub-module also uses historical data to accurately locate the capacitance value.

[0070] Using the same historical index sequence X = [2, 1, 1, 3, 2, 1, 4], and the corresponding historical effective capacitance value sequence Y (capacitance value / pF) = [126, 138, 135, 115, 122, 136, 108]. Construct a linear model , and apply the same least squares process (including the same termination conditions) to calculate the parameters and , with the goal of minimizing . The calculated pF / index unit and pF.

[0071] Use the current index 1 to substitute into the capacitance model for prediction: . Check the predicted value Whether it is within the recommended range [130 pF, 140 pF] corresponding to Index 1. The result is within the range. The system adopts this predicted value as the instantaneous capacitance parameter. This sub-module finally generates the instantaneous capacitance parameter .

[0072] Please refer to Figure 5 , the link verification module includes: The noise spectrum calculation sub-module calls the instantaneous inductance parameter and the instantaneous capacitance parameter, obtains the noise time-domain signal sequence at the output end of the filter, applies the fast Fourier transform to convert the time-domain signal into frequency-domain components, and calculates the ratio of the square of the amplitude of the frequency-domain components to the frequency bandwidth to generate the noise power spectral density; The noise spectrum calculation sub-module receives the instantaneous inductance parameter calculated by the inductance and capacitance parameter generation sub-module . The system first loads these two parameter values into the physical tunable filter circuit to make it operate under the new L / C configuration.

[0073] Subsequently, the time-domain acquisition of the noise signal at the output end of the filter is immediately started. Set the acquisition duration , and adopt a high sampling rate . Selecting 20 ms is to obtain enough signal cycles to ensure the frequency resolution, and a sampling rate of 10 MHz is to be able to analyze the noise components up to 5 MHz (Nyquist theorem), covering the wideband noise that the system may be concerned about. A total of discrete noise voltage sample points are obtained in this acquisition.

[0074] After obtaining the complete time-domain noise sequence, apply the fast Fourier transform (FFT) algorithm to it. The FFT efficiently converts these 200,000 time-domain sample points into a frequency-domain representation. Since the input is a real signal, the output of the FFT is a sequence containing complex frequency-domain coefficients. Each coefficient corresponds to a specific frequency . Therefore, the frequency-domain analysis covers the range from 0 Hz to .

[0075] Next, calculate the square of the magnitude of each frequency-domain coefficient . This value represents the noise power within the frequency interval with a center frequency of and a width of . To obtain the normalized noise power spectral density (PSD), divide the power at each frequency point by the frequency resolution . The calculation formula is . The unit of the calculation result is . By performing this calculation for all from 0 to N / 2, a series of noise power spectral density values at discrete frequency points are obtained. This sub-module finally generates this noise power spectral density curve containing (frequency, PSD) data points for subsequent analysis.

[0076] The threshold verification sub-module, based on the noise power spectral density, locates the frequency point at which the amplitude in the power spectrum decays to 70.7% of the maximum amplitude as the 3dB attenuation point, extracts the preset noise suppression bandwidth threshold, calculates the absolute difference between the attenuation point and the threshold, compares the difference with the threshold allowable range, determines whether the difference exceeds the upper limit of the range, and generates a difference verification result; The threshold allowable range is calibrated as [85%, 115%] through spectral analysis experiments on 500 groups of noise samples; The threshold verification sub-module receives the noise power spectral density data generated by the noise spectrum calculation sub-module . Its task is to evaluate whether the actual bandwidth of the filter under the current L / C parameter configuration meets the expectation. First, search and locate the maximum value of the power spectral density on the curve . This peak usually appears near the resonant frequency of the filter .

[0077] Find the peak and its corresponding center frequency . After that, starting from , search in the low-frequency and high-frequency directions respectively to find the frequency points at which the power spectral density value first drops to (i.e., the power decays by 3dB). Denote the 3dB point on the low-frequency side as , and the 3dB point on the high-frequency side as . By searching the calculated data, locate and .

[0078] Calculate the difference between these two 3dB frequency points to obtain the actual measured bandwidth of the filter . .

[0079] Next, the system retrieves the preset target noise suppression bandwidth threshold , which is provided (or directly set) by the bandwidth threshold matching sub-module. In this example . Then, compare the measured bandwidth with the target bandwidth . Calculate their consistency, usually measured by a ratio: .

[0080] The calculated ratio (ie 100%) is compared with the preset bandwidth allowable range. The allowable range is [85%, 115%], which is determined by statistical analysis experiments on the noise spectrum of 500 groups of samples of different working conditions and different devices. Statistical results show that when the bandwidth fluctuates within this range, the key performance indicators of the filter (such as selectivity and out-of-band suppression) can still meet the system requirements. Setting process: Calculate the distribution of the ratio of the measured bandwidth to the target bandwidth of 500 samples, take the 7.5th percentile and the 92.5th percentile as the lower and upper limits of the interval (approximately covering the middle 85% of the samples), and get [85%, 115%]. The judgment conditions are: .because , the condition is satisfied. Therefore, it is determined that the measurement bandwidth is within the allowable deviation range. This submodule finally generates the difference verification result: "Not exceeded".

[0081] The parameter validation submodule calls the difference verification result and executes the branch logic according to the over-limit status. If the limit is exceeded, the current parameter is cleared and the rollback instruction is triggered. If the limit is not exceeded, the parameter validation status is marked to generate the effective inductance parameter and the effective capacitance parameter.

[0082] The parameter validation submodule receives the difference verification result transmitted by the threshold verification submodule, and the current result is "within the limit". Subsequent operations are performed according to the internal control logic.

[0083] The logic states that if the received result is "out of limit" (meaning that the measurement bandwidth It fell on [85%,115%] The system will determine the instantaneous inductance parameters generated by the current calculation. and instantaneous capacitance parameters The filter fails to achieve the expected performance indicators. At this point, the system will discard this set of parameters, not apply them to the hardware, and perform a rollback operation: restore the last successful and still in use L / C parameter pair, and may trigger the parameter optimization algorithm to recalculate or issue an alarm.

[0084] However, since the current calibration result is "within the limit", the system executes the normal validation process. and instantaneous capacitance parameters The system then issues a command to write these two values ​​into the registers of the tunable filter hardware through the control interface, so that it can physically adjust the internal inductor and capacitor components (or equivalent circuits), so that the filter can operate according to the new parameters. This submodule finally generates the inductance and capacitance parameters that represent the actual operation in the current system and have been verified to be effective: effective inductance parameters And effective capacitance parameters These parameters will remain valid until the next successful parameter update process is completed.

[0085] A communication line noise filtering method, which is executed based on the above communication line noise filtering system, includes the following steps: S1: Respectively collect the real-time temperature values of the inductor node and the capacitor node through a dual-channel temperature sensor, input the two groups of temperature values into a piecewise linear regression model to calculate the temperature deviation gradient of the resonance frequency, and generate a temperature deviation gradient value; S2: Query a preset resonance frequency reference table based on the temperature deviation gradient value, match the noise suppression bandwidth threshold corresponding to the current temperature, and at the same time detect the input voltage amplitude volatility. Invoke the Kalman filtering algorithm to perform prediction processing on the voltage sampling sequence to generate a voltage fluctuation prediction value; S3: Locate the interval index in the inductor-capacitor parameter mapping table according to the voltage fluctuation prediction value, perform a linear fitting operation of the least squares method on the inductor value to generate an instantaneous inductor parameter, and synchronously perform the same operation on the capacitor value to generate an instantaneous capacitor parameter; S4: For the instantaneous inductor parameter and the instantaneous capacitor parameter, calculate the noise power spectral density at the output end of the filter through fast Fourier transform, compare the difference between the 3dB attenuation point of the power spectral density and the noise suppression bandwidth threshold. If the difference exceeds the preset range, trigger a parameter rollback instruction to regenerate the parameters, otherwise output the effective inductor parameter and the effective capacitor parameter to the filter control unit.

[0086] The above is only a preferred embodiment of the present invention, and it is not intended to limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A communication line noise filtering system, characterized in that, The system includes: A temperature coupling module, which is used to collect the real-time temperature value of the inductor node and the real-time temperature value of the capacitor node through a dual-channel thermosensor, input both into a piecewise linear regression model to calculate the deviation gradient of temperature on the resonance frequency, generate a temperature deviation gradient value, and transfer the temperature deviation gradient value and a preset resonance frequency reference table to a voltage response module; A voltage response module, which is used to receive the temperature deviation gradient value and the resonance frequency reference table, match the noise suppression bandwidth threshold corresponding to the current temperature based on the resonance frequency reference table, detect the input voltage amplitude volatility, call a Kalman filtering algorithm to process the voltage sampling sequence to generate a voltage fluctuation prediction value, and transfer it to a parameter joint adjustment module; A parameter joint adjustment module, which is used to receive the voltage fluctuation prediction value, extract an interval index from a preset inductor-capacitor parameter mapping table according to the voltage fluctuation prediction value, perform a linear fitting of the least squares method on the inductor value to generate an instant inductor parameter, perform the same operation on the capacitor value to generate an instant capacitor parameter, and transfer the instant inductor parameter and the instant capacitor parameter to a link verification module.

2. The communication line noise filtering system according to claim 1, characterized in that, The temperature deviation gradient value includes a temperature measurement error correction factor, a resonance frequency change compensation coefficient, and a node thermal response characteristic parameter. The voltage fluctuation prediction value includes a voltage stability index, a voltage fluctuation trend parameter, and a voltage noise pre-estimation value. The instant inductor parameter includes a dynamic inductor adjustment coefficient, a real-time temperature drift calibration amount, and an inductor tolerance correction amount. The instant capacitor parameter includes a dynamic capacitor adjustment coefficient, a real-time frequency drift calibration amount, and a capacitor tolerance correction amount.

3. The communication line noise filtering system according to claim 2, characterized in that, The piecewise linear regression model dynamically divides piecewise nodes based on the gradient change rate of the inductor and capacitor temperature intervals, and sets weight coefficients for linear regression within multiple intervals; The Kalman filtering algorithm constructs a state equation through an ambient temperature interference correction factor, a voltage attenuation dynamic coefficient, and a voltage sequence dispersion index; The inductor-capacitor parameter mapping table divides the upper and lower limits of the fluctuation interval based on 1.2 - 1.5 times the average value of the voltage volatility.

4. The communication line noise filtering system according to claim 3, characterized in that, The temperature coupling module includes: A temperature acquisition sub-module detects the temperature sensor signals of the inductor node and the capacitor node, uses a dual-channel signal synchronous acquisition circuit, sets a signal sampling interval and a noise filtering threshold, performs amplitude calibration and timing alignment on the original signals, respectively records the inductor temperature value and the capacitor temperature value, establishes two groups of real-time temperature time series data with synchronized timestamps, and generates dual-channel temperature values; A regression processing sub-module, based on the dual-channel temperature values, divides the piecewise nodes of the inductor and capacitor temperature intervals, sets weight coefficients for linear regression within multiple intervals, inputs the inductor temperature value as an independent variable and the capacitor temperature value as a covariate into the model, calculates the slope change amount of temperature and resonance frequency within multiple piecewise intervals, extracts a set of slope parameters for the piecewise intervals, and generates a regression coefficient set; The piecewise nodes are dynamically adjusted when the gradient change rate of adjacent temperature sampling points exceeds 0.5 °C / s. The gradient calculation sub-module calls the set of regression coefficients, combines the segmented interval to which the current dual-channel temperature value belongs, selects the corresponding slope parameter, and multiplies the difference between the dual-channel temperature value and the reference frequency value by the slope parameter according to the temperature-frequency mapping relationship in the resonance frequency reference table, outputs the frequency offset caused by a unit temperature change, and generates a temperature deviation gradient value.

5. The communication line noise filtering system according to claim 4, characterized in that, The voltage response module includes: Based on the temperature deviation gradient value, the threshold matching sub-module calls the temperature-frequency mapping relationship in the resonance frequency reference table, performs cubic spline interpolation calculation on the current temperature value, matches the nearest neighbor reference frequency point, extracts the upper and lower calibration values of the noise suppression bandwidth corresponding to the frequency point, and generates a bandwidth threshold matching value; The cubic spline interpolation satisfies that the first derivative is continuous at the nodes and the second derivative error is less than 0.1 Hz / °C; The fluctuation detection sub-module detects the peak-valley difference of the amplitude of the input voltage signal, intercepts the sampling sequence with a fixed time window, calculates the sum of the absolute values of the amplitude differences between adjacent sampling points in the window, and divides the sum by the product of the time window length and the reference amplitude to generate a voltage volatility; The filtering prediction sub-module calls the bandwidth threshold matching value and the voltage volatility, and uses the formula: ； Performs an operation to obtain a voltage fluctuation prediction component, and superimposes the prediction covariance matrix in the Kalman filter state equation to generate a voltage fluctuation prediction value; Among them, represents the voltage fluctuation prediction component, represents the temperature deviation gradient value, represents the resonance frequency reference value, represents the noise suppression bandwidth threshold, represents the voltage volatility, represents the environmental temperature interference correction factor, which is obtained by calibrating the contribution degree of temperature interference to voltage fluctuation through experiments, represents the voltage attenuation dynamic coefficient, which is fitted by regression analysis of historical voltage data, represents the voltage sequence dispersion index, which is the ratio of the standard deviation to the mean of the voltage sampling sequence.

6. The communication line noise filtering system according to claim 5, characterized in that, The parameter joint adjustment module includes: The fluctuation interval indexing sub-module collects the voltage fluctuation prediction value, and according to the interval division rules defined in the preset inductance-capacitance parameter mapping table, compares the prediction value with the upper and lower limit values of the fluctuation interval in the mapping table item by item, determines the index code of the interval to which the prediction value belongs, and generates a fluctuation interval index; The inductance parameter generation sub-module calls the fluctuation interval index, extracts the set of inductance values corresponding to the index in the mapping table, constructs a linear equation system with the index sequence as the independent variable and the inductance value as the dependent variable, calculates the minimum value of the sum of the squared residuals of the equation system by the least squares method, and solves the slope and intercept parameters of the inductance value to generate an instantaneous inductance parameter; The iteration termination condition of the least squares method is that the sum of the squared residuals is less than 0.01 or the number of iterations exceeds 50 times; The capacitance parameter generation sub-module calls the fluctuation interval index, extracts the set of capacitance values corresponding to the index in the mapping table, and based on the linear relationship between the index sequence and the capacitance value, applies the same least squares method iteration process as the inductance parameter generation sub-module to calculate the slope and intercept parameters of the capacitance value, and generates an instantaneous capacitance parameter.

7. The communication line noise filtering system according to claim 6, characterized in that, The system further includes: The link verification module is used to receive the instantaneous inductance parameter and the instantaneous capacitance parameter, calculate the noise power spectral density at the output end of the filter through fast Fourier transform, compare the difference between the 3 dB attenuation point of the power spectral density and the noise suppression bandwidth threshold, and if the difference exceeds the threshold, trigger a parameter rollback instruction and feedback it to the parameter joint adjustment module to regenerate the parameters, output the effective inductance parameter and the effective capacitance parameter and load them to the filter control unit.

8. The communication line noise filtering system according to claim 7, characterized in that, The effective inductance parameter and the effective capacitance parameter include noise suppression accuracy indicators, filter performance verification results, and link signal integrity evaluation values; The difference threshold is calibrated as ±10% of the noise suppression bandwidth threshold through a spectrum analysis experiment.

9. The communication line noise filtering system according to claim 8, wherein The link verification module includes: The noise spectrum calculation sub-module calls the instant inductor parameter and the instant capacitor parameter, obtains the noise time-domain signal sequence at the output end of the filter, converts the time-domain signal into frequency-domain components by applying the fast Fourier transform, calculates the ratio of the square of the amplitude of the frequency-domain components to the frequency bandwidth, and generates the noise power spectral density; The threshold verification sub-module, based on the noise power spectral density, locates the frequency point at which the amplitude in the power spectrum decays to 70.7% of the maximum amplitude as the 3dB attenuation point, extracts the preset noise suppression bandwidth threshold, calculates the absolute difference between the attenuation point and the threshold, compares the difference with the threshold allowable range, determines whether the difference exceeds the upper limit of the range, and generates a difference verification result; The threshold allowable range is calibrated as [85%, 115%] through a spectrum analysis experiment of 500 groups of noise samples; The parameter effective sub-module calls the difference verification result, executes branch logic according to the over-limit state. If it is over-limit, it clears the current parameters and triggers a rollback instruction. If it is not over-limit, it marks the parameter effective state and generates effective inductor parameters and effective capacitor parameters.

10. A communication line noise filtering method, wherein The method is used to implement the communication line noise filtering system according to any one of claims 1-9, and includes the following steps: S1: Respectively collect the real-time temperature values of the inductor node and the capacitor node through a dual-channel temperature sensor, input the two groups of temperature values into a piecewise linear regression model to calculate the temperature deviation gradient of the resonant frequency, and generate a temperature deviation gradient value; S2: Based on the temperature deviation gradient value, query a preset resonant frequency reference table to match the noise suppression bandwidth threshold corresponding to the current temperature. At the same time, detect the input voltage amplitude volatility, and call the Kalman filter algorithm to perform prediction processing on the voltage sampling sequence to generate a voltage fluctuation prediction value; S3: Locate the interval index in the inductor-capacitor parameter mapping table according to the voltage fluctuation prediction value, perform a linear fitting operation of the least squares method on the inductor value to generate an instant inductor parameter, and synchronously perform the same operation on the capacitor value to generate an instant capacitor parameter; S4: For the instant inductor parameter and the instant capacitor parameter, calculate the noise power spectral density at the output end of the filter through the fast Fourier transform, compare the difference between the 3dB attenuation point of the power spectral density and the noise suppression bandwidth threshold. If the difference exceeds the preset range, trigger a parameter rollback instruction to regenerate the parameters. Otherwise, output the effective inductor parameters and effective capacitor parameters to the filter control unit.

Images