An intelligent dynamic response phase-locked loop filtering method and system

Abstract

This application relates to an intelligent dynamic response phase-locked loop (PLL) filtering method and system, belonging to the field of PLL technology. The method includes: performing segmented collaborative processing on the PLL output signal to detect the noise characteristics of the output signal in real time, wherein the noise characteristics include at least frequency domain noise distribution information of the output signal; dynamically matching the optimal filter bandwidth under the current noise environment based on the noise characteristics; and adjusting the actual bandwidth of the filter in the PLL based on the optimal filter bandwidth. This application can dynamically adjust its bandwidth according to the real-time signal and noise environment to obtain optimal filtering performance under different noise conditions.

Description

Technical Field

[0001] This application relates to the field of phase-locked loop (PLL) technology, and in particular to a smart dynamic response PLL filtering method and system. Background Technology

[0002] Phase-locked loops (PLLs), as the core module for frequency synthesis and clock synchronization, are widely used in communications, radar, measurement and testing, and other fields. The filters in a PLL are used to suppress noise and interference in the input signal, and their bandwidth selection directly affects the PLL's lock-in time, phase noise suppression capability, and system stability.

[0003] Traditional phase-locked loop (PLL) filters often employ a fixed bandwidth design, meaning a compromise filter bandwidth is pre-defined based on the application scenario. For example, in wireless communication receivers, designers select a fixed loop filter bandwidth based on the expected signal characteristics and noise environment to achieve a balance between lock-in time and phase noise.

[0004] However, in practical applications, the signal environment in which phase-locked loops (PLLs) operate is often complex and variable. When signal noise increases, a fixed-bandwidth filter, if its bandwidth is too wide, cannot effectively suppress noise, leading to deterioration of phase noise; if its bandwidth is too narrow, it cannot respond in time when the signal changes rapidly, resulting in prolonged lock-in time or even loss of lock-in. Due to the lack of perception of real-time noise environment and adaptive adjustment mechanism, traditional fixed-bandwidth PLL filters struggle to maintain optimal filtering performance in dynamically changing signal environments, a problem that urgently needs to be solved. Summary of the Invention

[0005] In order to dynamically adjust its bandwidth according to the real-time signal and noise environment and obtain the optimal filtering effect under different noise conditions, this application provides an intelligent dynamic response phase-locked loop filtering method and system.

[0006] Firstly, this application provides an intelligent dynamic response phase-locked loop filtering method, including:

[0007] The phase-locked loop output signal is processed in segments to detect the noise characteristics of the output signal in real time. The noise characteristics include at least the frequency domain noise distribution information of the output signal.

[0008] Based on the noise characteristics, the optimal filter bandwidth is dynamically matched under the current noise environment;

[0009] Adjust the actual bandwidth of the filter in the phase-locked loop according to the optimal filter bandwidth.

[0010] Optionally, the method further includes:

[0011] Real-time monitoring of the ambient temperature of the phase-locked loop's environment, combined with historical temperature data, to predict temperature change trends;

[0012] Based on the temperature change trend, the compensation amount of the control parameters of the voltage-controlled oscillator is determined, and the control parameters of the voltage-controlled oscillator are adjusted according to the compensation amount to compensate for the frequency drift caused by temperature changes.

[0013] Optionally, the method further includes:

[0014] Detect whether the phase-locked loop has entered a stable locked state, which means that the phase-locked loop has established a stable frequency and phase tracking state;

[0015] When the phase-locked loop enters a stable locked state, at least one high-power module in the phase-locked loop is turned off, so that the phase-locked loop enters a low-power mode; the high-power module refers to the circuit unit in the phase-locked loop whose power consumption is higher than a preset value in normal operation mode.

[0016] In the low-power mode, the phase error of the phase-locked loop is monitored;

[0017] When the phase error is detected to exceed the preset threshold, the high-power module that was shut down is restarted.

[0018] Optionally, the step of performing segmented collaborative processing on the phase-locked loop output signal to detect the noise characteristics of the output signal in real time includes:

[0019] The phase-locked loop output signal is processed by overlapping segmentation, and there is an overlapping area between adjacent signal segments;

[0020] The signal change rate of the output signal is detected, and the length of the overlapping region and / or the fusion weight of the overlapping region are dynamically adjusted according to the signal change rate.

[0021] Spectral analysis is performed on each signal segment in parallel to obtain the frequency domain noise distribution information of each signal segment; based on the adjusted fusion weights, the spectral analysis results of the overlapping region are weighted and fused to obtain the frequency domain noise distribution information of the overlapping region.

[0022] Optionally, the method further includes:

[0023] Obtain the noise feature sequence of each signal segment;

[0024] Analyze the correlation of noise features between adjacent signal segments, and determine the influence features corresponding to each signal segment based on the correlation of noise features; the influence features are used to characterize the range and / or degree of influence of the corresponding signal segment on subsequent signal segments;

[0025] The step of dynamically matching the optimal filter bandwidth under the current noise environment based on the noise characteristics includes:

[0026] Based on the noise characteristics of the current signal segment and the corresponding influence characteristics, predict the noise evolution trend of the signal segment at subsequent time moments;

[0027] Based on the noise evolution trend, dynamically match the optimal filter bandwidth sequence for the current and subsequent time points;

[0028] The step of adjusting the actual bandwidth of the filter in the phase-locked loop according to the optimal filter bandwidth includes:

[0029] Based on the optimal filter bandwidth sequence, adjust the actual bandwidth of the filter in the phase-locked loop at the current and subsequent times.

[0030] Optionally, adjusting the actual bandwidth of the filter in the phase-locked loop at the current and subsequent times according to the optimal filter bandwidth sequence includes:

[0031] Analyze the direction and intensity of influence transmission between different signal segments;

[0032] Based on the direction of the influence propagation, determine the optimal bandwidth adjustment timing for the current and subsequent moments;

[0033] The execution step size for bandwidth adjustment is determined based on the intensity of the influence transmission, wherein the greater the intensity of the influence transmission, the smaller the execution step size;

[0034] According to the optimal filter bandwidth sequence and execution step size, at the optimal bandwidth adjustment time, the actual bandwidth of the filter in the phase-locked loop is adjusted.

[0035] Optionally, predicting the noise evolution trend of signal segments at subsequent times based on the noise characteristics of the current signal segment and the corresponding influence characteristics includes:

[0036] Based on the noise characteristics of the current signal segment and the corresponding influence characteristics, predict the noise evolution trend of the signal segment at subsequent time moments;

[0037] The noise characteristics of the current signal segment are compared with the noise characteristics of the previous signal segment to obtain the noise variation amplitude;

[0038] The final noise evolution trend is obtained by correcting the noise evolution trend based on the noise change amplitude, wherein the larger the noise change amplitude, the greater the correction weight.

[0039] Secondly, this application provides an intelligent dynamic response phase-locked loop filter system, including,

[0040] The signal monitoring module is used to perform segmented collaborative processing on the phase-locked loop output signal to detect the noise characteristics of the output signal in real time. The noise characteristics include at least the frequency domain noise distribution information of the output signal.

[0041] A bandwidth matching module is used to dynamically match the optimal filter bandwidth under the current noise environment based on the noise characteristics.

[0042] The bandwidth adjustment module is used to adjust the actual bandwidth of the filter in the phase-locked loop according to the optimal filter bandwidth.

[0043] Thirdly, this application provides an intelligent dynamic response phase-locked loop filter device, including a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as described in any of the first aspects.

[0044] Fourthly, this application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described in any of the first aspects.

[0045] In summary, this application includes at least one of the following beneficial technical effects:

[0046] In this application, by performing segmented collaborative processing on the phase-locked loop (PLL) output signal, the frequency domain noise distribution information of the signal can be detected in real time. Based on this noise characteristic, the optimal filter bandwidth under the current noise environment is dynamically matched, thereby adjusting the actual bandwidth of the filter. This allows the PLL filter to no longer use fixed bandwidth parameters, but instead automatically adjust according to the real-time noise environment, thus maintaining good filtering performance under different noise conditions.

[0047] Furthermore, by monitoring ambient temperature in real time and combining it with historical temperature data to predict temperature change trends, the compensation amount of the voltage-controlled oscillator (VCO) control parameters can be determined in advance and adjusted to counteract frequency drift caused by temperature changes. This allows the phase-locked loop (PLL) to maintain a stable output frequency even when ambient temperature changes, reducing the impact of temperature on system performance.

[0048] Furthermore, by detecting whether the phase-locked loop (PLL) has entered a stable locked state, and shutting down the high-power module to enter a low-power mode after stabilization, while continuously monitoring phase error in low-power mode, the shut-down module is immediately woken up if a risk of loss of lock is detected. This allows the PLL to significantly reduce power consumption while maintaining stable tracking, making it particularly suitable for applications that require long-term operation and are power-sensitive. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating an intelligent dynamic response phase-locked loop filtering method disclosed in an embodiment of this application.

[0051] Figure 2 This is a structural block diagram of an intelligent dynamic response phase-locked loop filter system disclosed in an embodiment of this application. Detailed Implementation

[0052] The following is in conjunction with the appendix Figure 1-2 This application will be described in further detail.

[0053] This application discloses an intelligent dynamic response phase-locked loop filtering method (hereinafter referred to as the intelligent control method), and the corresponding execution subject is an intelligent dynamic response phase-locked loop filtering system (hereinafter referred to as the intelligent control system). The following will refer to... Figure 1 This section elaborates on the detailed execution process of intelligent control methods by the intelligent control system.

[0054] S101 performs segmented collaborative processing on the phase-locked loop output signal to detect the noise characteristics of the output signal in real time. The noise characteristics include at least the frequency domain noise distribution information of the output signal.

[0055] S102 dynamically matches the optimal filter bandwidth under the current noise environment based on noise characteristics.

[0056] S103, adjust the actual bandwidth of the filter in the phase-locked loop according to the optimal filter bandwidth.

[0057] S104 monitors the ambient temperature of the phase-locked loop in real time, combines historical temperature data to predict temperature change trends, determines the compensation amount of the control parameters of the voltage-controlled oscillator based on the temperature change trend, and adjusts the control parameters of the voltage-controlled oscillator according to the compensation amount to compensate for frequency drift caused by temperature changes.

[0058] S105, detect whether the phase-locked loop (PLL) has entered a stable locked state. The stable locked state means that the PLL has established a stable frequency and phase tracking state. When the PLL enters the stable locked state, shut down at least one high-power module in the PLL, so that the PLL enters a low-power mode. The high-power module refers to the circuit unit in the PLL whose power consumption is higher than a preset value in the normal operating mode. In the low-power mode, monitor the phase error of the PLL. When the phase error is detected to exceed the preset threshold, restart the shut-down high-power module.

[0059] In implementation, the intelligent control system first performs segmented collaborative processing on the phase-locked loop (PLL) output signal to detect its noise characteristics in real time. Segmented collaborative processing refers to dividing the PLL output signal into multiple time-series consecutive signal segments and then performing collaborative analysis on these segments. Through this collaborative analysis, the dynamic evolution of signal noise over time can be captured, laying the foundation for subsequent dynamic matching of loan data.

[0060] In practice, the output signal of the phase-locked loop (PLL) is first acquired. This output signal can be the output signal of the voltage-controlled oscillator (VCO) in the PLL, or it can be the feedback signal processed by a frequency divider. In this embodiment, the output signal of the VCO is preferably used as the analysis object because this signal directly reflects the output quality of the PLL.

[0061] Next, the output signal is segmented along the time axis. Segmentation can be done using equal-length segments or variable-length segments depending on the signal characteristics. For example, the continuous output signal can be divided into a series of time-sequentially consecutive signal segments Q1, Q2, Q3…Qn. Each signal segment covers a time interval of [t0, t0+L), [t0+L, t0+2L), [t0+2L, t0+3L), and so on. The choice of the time window length L needs to strike a balance between time resolution and frequency resolution, because a large L will reduce the frequency resolution of the spectrum analysis, while an excessively long L will make it difficult to capture rapidly changing noise characteristics. Preferably, the value of L ranges from 1μs to 100μs, and can be configured according to the actual application scenario.

[0062] After segmentation processing, the intelligent control system performs parallel spectrum analysis on each signal segment to obtain the frequency domain noise distribution information of each segment. Parallel processing is a key technical means for achieving real-time detection in this application; that is, performing spectrum analysis on multiple signal segments simultaneously, rather than processing them sequentially one by one. Parallel processing significantly reduces the noise detection delay, enabling the intelligent control system to respond quickly to changes in the noise environment. The specific implementation of spectrum analysis employs the Fast Fourier Transform (FFT) algorithm. The FFT algorithm can convert a time-domain signal into a frequency-domain signal, thereby obtaining the energy distribution of the signal at different frequency components. The FFT algorithm itself is a known fast algorithm, and this application uses it as the basic tool for spectrum analysis.

[0063] To further improve the speed of spectrum analysis and achieve real-time analysis performance within 2μs, this application includes a hardware acceleration module to accelerate the execution of FFT operations. This hardware acceleration module is integrated into the phase-locked loop chip, and its functions include:

[0064] Input interface: Receives signal segment data from the segmentation processing module;

[0065] The computation unit performs the core operations in FFT operations, including butterfly operations, complex multiplication, and complex addition.

[0066] Storage unit: Temporarily stores intermediate operation results and rotation factors;

[0067] Control unit: Schedules the coordinated work of the computing unit and the storage unit according to the progress of the FFT operation;

[0068] Output interface: Outputs the calculated frequency domain data for subsequent modules to perform power spectral density calculations.

[0069] In its implementation, the hardware acceleration module employs a pipelined architecture: while the current data point is undergoing butterfly operations, subsequent data points are already being read and preprocessed, enabling parallel processing of multiple data points. Simultaneously, the hardware acceleration module can internally configure multiple parallel butterfly operation units to further enhance processing capabilities. For an N-point FFT operation, the hardware acceleration module can complete all calculations within N clock cycles, significantly faster than the method of a general-purpose processor executing software code. Through this hardware acceleration module, the system can quickly calculate the power spectral density (PSD) or amplitude spectrum for each signal segment. This spectral information represents the frequency domain noise distribution information for that time segment, clearly reflecting which frequency components in the signal are subject to noise interference and the magnitude of the noise energy. For example, if the output signal is interfered with at a specific frequency, a high energy peak will appear on the spectral line corresponding to that frequency; if broadband noise exists, the spectral lines will exhibit a certain energy level across the entire frequency band. Through optimization of the hardware acceleration module, this application can shorten the noise spectrum analysis time to less than 2 μs, meaning the entire process from signal input to obtaining complete frequency domain noise distribution information takes no more than 2 microseconds. This performance metric enables the system to capture rapid changes in noise with extremely high temporal resolution, providing a real-time and accurate basis for subsequent dynamic bandwidth adjustments.

[0070] After obtaining the frequency domain noise distribution information of each signal segment, the noise characteristics of the output signal at the current moment can be obtained. It should be noted that the noise characteristics described in this application include at least the aforementioned frequency domain noise distribution information, but in other embodiments, other noise-related parameters such as the overall noise level may also be included. The overall noise level refers to the total intensity of noise energy of the output signal over a wide frequency band, used to macroscopically evaluate the noise intensity of the signal. In a practical system, the overall noise level can be expressed in one of the following specific forms: RMS value (the root mean square voltage value of the output signal within a preset frequency band, in volts (V) or millivolts (mV)); power value (the average power of the output signal within a preset frequency band, in watts (W) or decibel-milliwatts (dBm)); quantization code value (a digital quantity obtained after analog-to-digital conversion, proportional to the noise energy). The overall noise level can be obtained using a broadband RMS detector, a circuit module capable of directly measuring the true effective value (RMS) of a signal. Its working principle is as follows: First, the phase-locked loop output signal is input to the broadband RMS detector. This broadband RMS detector features a wide bandwidth, which should cover the noise frequency range to be monitored. For example, if noise needs to be monitored in the 1MHz to 100MHz frequency band, the bandwidth of the RMS detector should at least cover this band. A broadband RMS detector typically contains the following core units: a squarer (squaring the instantaneous voltage of the input signal to obtain the instantaneous power), an integrator (integrating and averaging the squared signal to obtain the average power), a square root (taking the square root of the average power to obtain the RMS value), and an output buffer (outputting the RMS value to subsequent processing units). In specific implementations, the broadband RMS detector can be implemented using analog or digital circuits. Analog RMS detectors have the advantages of fast response speed and wide bandwidth, making them suitable for high-frequency signals; digital RMS detectors, on the other hand, use a high-speed ADC to sample the signal and calculate the RMS value in the digital domain, offering advantages of high accuracy and strong configurability. Those skilled in the art can choose the appropriate implementation method based on the actual application scenario.

[0071] Next, in step S102, the optimal filtering bandwidth is dynamically matched to the current noise environment based on the detected noise characteristics. The optimal filtering bandwidth refers to the loop filter bandwidth that enables the phase-locked loop to achieve its best overall performance under the current noise environment. The selection of this bandwidth requires a balance between phase noise suppression capability and dynamic response speed: too wide a bandwidth results in a fast loop response but introduces more out-of-band noise, leading to deterioration of the limiting noise; too narrow a bandwidth provides strong noise suppression but results in a slow loop response, making it difficult to track rapidly changing signals. Therefore, the optimal filtering bandwidth should be a dynamically determined value based on the current noise characteristics, achieving the best balance between the two. The specific implementation of dynamically matching the optimal filtering bandwidth in this application can employ a preset noise-bandwidth mapping table, which can be stored in a digital state machine. This mapping table is established in advance by human experimentation during the intelligent control system design phase, establishing a correspondence between noise characteristics and the optimal bandwidth. This mapping table records the optimal bandwidth value to be used under different noise conditions (such as in-band noise of different intensities, interference of different frequencies, etc.). In actual operation, after obtaining the current noise characteristics through step S101, the digital state machine can quickly obtain the corresponding optimal filter bandwidth by looking up a table and drive the programmable transconductance amplifier to perform bandwidth adjustment. In a more preferred embodiment, this mapping relationship can be multi-dimensional, considering multiple noise characteristic parameters simultaneously. For example, based on the distribution characteristics of the noise power spectral density, the minimum bandwidth required for in-band noise suppression and the maximum bandwidth required for out-of-band noise suppression can be determined separately, and then the optimal bandwidth value can be selected within this range.

[0072] Finally, in step SS103, the intelligent control system adjusts the actual bandwidth of the filter in the phase-locked loop (PLL) based on the determined optimal filter bandwidth. The loop filter in a PLL typically employs a programmable or configurable design to dynamically adjust its cutoff frequency. For example, for analog loop filters, bandwidth adjustment can be achieved using a programmable transconductance amplifier (OTA) or a programmable resistor-capacitor array. When the filter bandwidth needs adjustment, a corresponding control code is sent to the OTA to change its transconductance value, thereby altering the filter's equivalent resistance and cutoff frequency. For digital loop filters, their frequency response characteristics can be adjusted directly by modifying the digital filter coefficients. During the adjustment process, the intelligent control system calculates the required filter parameters based on the optimal bandwidth value determined in step S102, and then writes these parameters into the filter configuration register via the control interface, completing the actual bandwidth adjustment. After adjustment, the PLL will continue operating with the new filter bandwidth, thereby achieving optimal phase noise suppression and dynamic response performance under the current noise environment.

[0073] Furthermore, the intelligent control method of this application also includes a temperature compensation mechanism to address the impact of ambient temperature changes on the performance of the phase-locked loop. This mechanism is implemented through step S104.

[0074] First, the intelligent control system monitors the ambient temperature of the environment in which the phase-locked loop (PLL) is located in real time. This can be achieved by integrating a temperature sensor inside the PLL chip or by deploying a temperature detection circuit near the chip. The temperature sensor converts the ambient temperature into an electrical signal (such as voltage, current, or digital code value) for subsequent processing. Temperature monitoring can be continuous or performed according to a preset sampling period, such as collecting temperature data every 100ms.

[0075] Secondly, the intelligent control system combines historical temperature data to predict temperature change trends. A simple current temperature value only reflects the instantaneous temperature situation but cannot predict how the temperature will change. To compensate for temperature-induced frequency drift in advance, it is necessary to understand the temperature change trend—whether the temperature is rising or falling, the rate of change, and the likely future temperature value.

[0076] Predicting temperature change trends can be achieved using various methods. A simple, exemplary method is linear extrapolation: recording temperature values ​​over multiple consecutive moments, calculating the rate of temperature change with time, and then assuming the temperature will continue to change at that rate to predict the temperature at future moments. For example, if the current temperature is T0 and the previous temperature was T-1, then the rate of temperature change ΔT / Δt = (T0 - T-1) / Δt, and the predicted temperature Δt moments later is T_pred = T0 + (ΔT / Δt) × Δt.

[0077] In a preferred embodiment, machine learning algorithms can be used for temperature trend prediction. For example, historical temperature data over a longer period can be collected to train a lightweight neural network or support vector machine model to predict temperature change trajectories. Considering the limited computational resources of the phase-locked loop chip, simplified time series prediction models, such as ARIMA (Autoregressive Integral Moving Average) or exponential smoothing, can be used. These models can predict future temperature change trends based on patterns in historical data.

[0078] Next, the intelligent control system determines the compensation amount for the control parameters of the voltage-controlled oscillator (VCO) based on the predicted temperature change trend. The VCO is the core module in the phase-locked loop (PLL) that generates the output frequency, and its oscillation frequency drifts due to temperature. Different types of VCOs have different temperature sensitivities: VCOs based on LC resonant cavities have relatively small temperature coefficients, while VCOs based on ring oscillators have larger temperature coefficients. Regardless of the type, temperature changes will cause changes in the VCO output frequency, thereby increasing the phase error of the PLL output signal.

[0079] The effect of temperature on a VCO is typically described by its temperature characteristic curve. This curve represents the amount of adjustment required to maintain the same output frequency at different temperatures. The VCO's temperature characteristic curve or temperature coefficient can be obtained through experimental measurements or from the chip datasheet. Based on the temperature trend predicted by S102 (e.g., the difference ΔT between the predicted future temperature T_pred and the current temperature T_current), and combined with the VCO's temperature coefficient K_temp (unit: Hz / ℃ or V / ℃), the required control voltage compensation ΔV_comp = K_temp × ΔT to offset the temperature-induced frequency drift can be calculated. If the VCO is digitally controlled (DCO), the compensation is expressed as the adjustment value ΔCode of the control word.

[0080] Finally, the intelligent control system adjusts the control parameters of the voltage-controlled oscillator (VCO) based on the determined compensation amount. For analog VCOs, adjusting the control parameters involves superimposing a compensation voltage ΔV_comp onto the original control voltage, ensuring the VCO maintains its original output frequency after temperature changes. For digitally controlled oscillators (DCOs), the adjustment involves modifying the frequency control word, increasing or decreasing ΔCode. Adjustments can be made in advance before temperature changes occur (based on predicted trends) or in real-time after temperature changes occur. Advance adjustment is preferred to eliminate transient frequency fluctuations caused by temperature changes.

[0081] Furthermore, the intelligent control method of this application also includes a power consumption management mechanism to reduce system power consumption while ensuring performance. This mechanism is implemented through S105. First, the intelligent control system detects whether the phase-locked loop (PLL) has entered a stable locked state. A stable locked state means that the PLL has established a stable frequency and phase tracking relationship, the frequency of the output signal is completely consistent with the reference signal, and the phase difference remains constant and approaches zero. In a practical system, the stable locked state can be determined by monitoring the phase error signal. For example, when the absolute value of the phase error is consistently less than a preset threshold (e.g., 0.01 radians) and the duration reaches a preset duration (e.g., 10 ms), it can be determined that the PLL has entered a stable locked state. The phase error here can be obtained through a phase detector or a time-to-digital converter (TDC).

[0082] Once the phase-locked loop (PLL) enters a stable locked state, the intelligent control system will shut down at least one high-power module within the PLL, putting the PLL into a low-power mode. A high-power module refers to a circuit unit in the PLL whose power consumption exceeds a preset value during normal operation. These modules are typically precision measurement circuits that require continuous high-speed operation, such as high-precision time-to-digital converters (TDCs). TDCs convert phase errors into digital signals; the higher their accuracy and sampling rate, the greater their power consumption. After the PLL enters a stable locked state, the phase error is small and stable, and the TDC is no longer needed for continuous high-precision measurement, so it can be shut down to save power.

[0083] Besides TDC (Transmission Controlled Controller), other high-power modules may include high-frequency dividers, high-speed frequency and phase detectors, etc. The intelligent control system can selectively shut down one or more high-power modules based on preset priorities or power consumption levels. Shutting down a module can be done by cutting off its clock source, disconnecting its power supply, or putting it into sleep mode.

[0084] After the phase-locked loop (PLL) enters low-power mode, the intelligent control system does not completely stop monitoring the phase error. Instead, it continuously monitors the PLL's phase error through a low-power monitoring circuit. This monitoring circuit can be a critical comparator circuit. The critical comparator circuit's function is to quickly determine whether the phase error exceeds the allowable range without needing to precisely measure the specific value of the phase error. When the phase error is small, the comparator outputs a low level (or a high level, depending on the design); when the phase error exceeds a preset threshold, the comparator output flips, generating a trigger signal. This trigger signal is used to wake up the high-power modules that have been shut down.

[0085] When the phase error exceeds the preset threshold, it indicates a risk of phase-locked loop (PLL) loss of lock, possibly due to sudden changes in the input signal, environmental interference, or other reasons. In this case, the intelligent control system immediately restarts the previously disabled high-power module, restoring the PLL to full-function operation. The wake-up process should be rapid enough to allow for timely and accurate measurement and adjustment of the phase error, preventing PLL loss of lock.

[0086] After resuming full-function mode, the phase-locked loop re-enters the normal dynamic adjustment process, adjusting the filter bandwidth in real time again via steps S101-S103, and performing temperature compensation via step S104. Once the phase error stabilizes again, the intelligent control system can re-enter low-power mode to achieve dynamic power consumption management.

[0087] Optionally, S101 specifically includes the following sub-steps:

[0088] S1011 performs overlapping segmentation processing on the phase-locked loop output signal, and there is an overlapping area between adjacent signal segments;

[0089] S1012, detect the signal change rate of the output signal, and dynamically adjust the length of the overlapping region and / or the fusion weight of the overlapping region according to the signal change rate;

[0090] S1013, perform parallel spectrum analysis on each signal segment to obtain the frequency domain noise distribution information of each signal segment; according to the adjusted fusion weight, perform weighted fusion processing on the spectrum analysis results of the overlapping region to obtain the frequency domain noise distribution information of the overlapping region.

[0091] The intelligent control method also includes the following steps:

[0092] S301, obtain the noise feature sequence of each signal segment.

[0093] S302, Analyze the correlation of noise characteristics between adjacent signal segments, and determine the influence characteristics corresponding to each signal segment based on the correlation of noise characteristics; the influence characteristics are used to characterize the range and / or degree of influence of the corresponding signal segment on subsequent signal segments.

[0094] S102 specifically includes the following sub-steps:

[0095] S1021, Based on the noise characteristics of the current signal segment and the corresponding influence characteristics, predict the noise evolution trend of the signal segment at subsequent time moments; compare the noise characteristics of the current signal segment with the noise characteristics of the previous signal segment to obtain the noise change amplitude; correct the noise evolution trend based on the noise change amplitude to obtain the final noise evolution trend, wherein the larger the noise change amplitude, the greater the correction weight.

[0096] S1022, dynamically matches the optimal filter bandwidth sequence for the current and subsequent moments based on the noise evolution trend.

[0097] S103 specifically includes the following sub-steps:

[0098] S1031, Analyze the direction and intensity of influence transmission between each signal segment;

[0099] S1032, Determine the optimal bandwidth adjustment timing for the current and subsequent moments based on the direction of influence propagation;

[0100] S1033, Determine the execution step size for bandwidth adjustment based on the influence transmission strength, wherein the greater the influence transmission strength, the smaller the execution step size;

[0101] S1034, according to the optimal filter bandwidth sequence and execution step size, adjusts the actual bandwidth of the filter in the phase-locked loop at the optimal bandwidth adjustment time.

[0102] In implementation, S1011-S1013 further refine S101. Specifically: First, the intelligent control system performs overlapping segmentation processing on the phase-locked loop output signal, with overlapping regions between adjacent signal segments. Overlapping segmentation processing means that when the output signal is divided into multiple continuous signal segments according to time, adjacent signal segments are not immediately connected, but rather have an overlapping interval of a certain length. For example, if the signal segment length is L, then the overlap length is M (0 < M < L). The time interval covered by the i-th signal segment is [t_i, t_i+L), and the time interval covered by the (i+1)-th segment is [t_i+LM, t_i+2L-M), with an overlap interval of [t_i+LM, t_i+L). The purpose of overlapping segmentation is to avoid signal loss due to signal truncation at segment boundaries, thereby improving the continuity and accuracy of spectrum analysis.

[0103] Next, the intelligent control system detects the rate of change of the output signal and dynamically adjusts the length of the overlapping region and / or the fusion weight of the overlapping region based on the rate of change. The rate of change of the signal is used to characterize the drastic change of the output signal over time and can be obtained in various ways. For example, the rate of change of the signal amplitude, the zero-crossing rate, or the rate of change of the noise features detected in S101 can be calculated. In this embodiment, the difference in frequency domain noise distribution information between adjacent signal segments is preferably used to characterize the rate of change of the signal. Specifically, for two adjacent signal segments, the power spectral density values ​​of the two at each frequency point are subtracted point by point, and then the sum of the squares or absolute values ​​of the differences at all frequency points is calculated. The result is the rate of change of the signal at the current moment. The larger the value, the more drastic the signal change; the smaller the value, the smoother the signal change. To improve robustness, the rate of change at multiple consecutive moments can be averaged. For example, the average of the most recent 3, 5, or 10 rate of change can be taken as the current rate of change used for decision-making, avoiding erroneous adjustments caused by a single abnormal fluctuation.

[0104] The overlap region length refers to the duration of the overlap region between two adjacent signal segments. In this embodiment, the adjustment range of the overlap region length is set between 10% and 50% of the signal segment length. A minimum overlap length of 10% ensures the basic continuity of the overlap analysis and avoids information loss at segment boundaries due to insufficient overlap. A maximum overlap length of 50% avoids computational redundancy caused by excessive overlap, while ensuring that each segment still has sufficient independent analysis area. The adjustment of the overlap region length is based on the signal change rate: the higher the signal change rate, the shorter the overlap region length; the lower the signal change rate, the longer the overlap region length. The specific adjustment method is as follows:

[0105] First, high and low rate of change thresholds are pre-set. These thresholds can be obtained through calibration of the intelligent control system before actual operation. For example, by measuring the statistical range of signal change rates in typical application scenarios, the upper limit of the statistical range can be used as the high threshold and the lower limit as the low threshold; alternatively, they can be dynamically updated based on historical data during system operation. When the detected signal change rate is lower than or equal to the low rate of change threshold, the signal is considered to be in a highly stable state, and the overlap region length is set to its maximum value, i.e., 50% of the total length of the signal segment. When the detected signal change rate is higher than or equal to the high rate of change threshold, the signal is considered to be in a state of rapid change, and the overlap region length is set to its minimum value, i.e., 10% of the total length of the signal segment. When the detected signal change rate is between the low and high rate of change thresholds, the overlap region length varies linearly between 10% and 50%: the closer the signal change rate is to the low rate of change threshold, the closer the overlap length is to 50%; the closer the signal change rate is to the high rate of change threshold, the closer the overlap length is to 10%. Specifically, the overlap region length is calculated as follows: First, calculate the difference between the current signal change rate and the low change rate threshold, then divide it by the difference between the high change rate threshold and the low change rate threshold to obtain a ratio between 0 and 1. This ratio represents the relative position of the current change rate within the range from the low threshold to the high threshold; the closer to 0, the closer to the low threshold, and the closer to 1, the closer to the high threshold. Then, multiply this ratio by the adjustable range of the overlap region length (i.e., 50% of the maximum overlap length minus 10% of the minimum overlap length, resulting in an adjustable range of 40%) to obtain an adjustment amount. Since a higher change rate results in a shorter overlap length, this adjustment amount needs to be subtracted from 50% of the maximum overlap length to obtain the percentage of the overlap length that should be set.

[0106] The fusion weight refers to the weight coefficient of the analysis result of the overlapping region in the final result when performing spectral analysis on the overlapping region. In this embodiment, the adjustment range of the fusion weight is set between 0.2 and 0.8. Too low a weight will weaken the smoothing effect of the overlapping region, potentially causing jumps at segment boundaries; too high a weight may lead to excessive dilution of the information in the current segment, reducing the real-time performance of the analysis. The fusion weight is adjusted based on the signal change rate, but in the opposite direction to the overlap length: the higher the signal change rate, the lower the fusion weight; the lower the signal change rate, the higher the fusion weight. The specific adjustment method is as follows:

[0107] The pre-set high and low rate of change thresholds are the same as those used in the overlap length adjustment. When the detected signal rate of change is lower than or equal to the low rate of change threshold, the signal is considered extremely stable, and the information in the overlapping area has high reference value; in this case, the fusion weight is set to the maximum value of 0.8. When the detected signal rate of change is higher than or equal to the high rate of change threshold, the signal is considered to be changing drastically, and the information in the overlapping area has high "staleness"; in this case, the fusion weight is set to the minimum value of 0.2. When the detected signal rate of change is between the low and high rate of change thresholds, the fusion weight changes linearly between 0.2 and 0.8: the closer the signal rate of change is to the low threshold, the closer the fusion weight is to 0.8; the closer the signal rate of change is to the high threshold, the closer the fusion weight is to 0.2. Similarly, the specific fusion weight is calculated as follows: first, calculate the difference between the current signal rate of change and the low rate of change threshold, then divide by the difference between the high and low rate of change thresholds to obtain a ratio value between 0 and 1. This ratio value represents the relative position of the current rate of change within the range from the low to the high threshold. Then, multiply this ratio by the adjustable range of the fusion weight (i.e., the maximum fusion weight 0.8 minus the minimum fusion weight 0.2, resulting in an adjustable range of 0.6) to obtain an adjustment amount. Since a higher rate of change results in a lower fusion weight, this adjustment amount needs to be subtracted from the maximum fusion weight 0.8 to obtain the fusion weight that should be set currently.

[0108] Next, the intelligent control system performs parallel spectrum analysis on each signal segment to obtain the frequency domain noise distribution information of each signal segment. As before, the spectrum analysis is specifically implemented using the Fast Fourier Transform (FFT) algorithm. To accelerate the FFT calculation speed, this embodiment uses a hardware acceleration module to implement the FFT calculation. This hardware acceleration module is integrated into the phase-locked loop chip, and its functional units include: an input interface for receiving time-domain data from each signal segment; a computation unit for performing butterfly operations, complex multiplication, and complex addition in the FFT calculation; a storage unit for temporarily storing intermediate calculation results and twiddle factors; a control unit for scheduling the collaborative work of the computation unit and the storage unit; and an output interface for outputting the calculated frequency domain data. The hardware acceleration module adopts a pipelined architecture, capable of completing N-point FFT calculations within N clock cycles, achieving fast spectrum analysis.

[0109] After obtaining the frequency domain noise distribution information of each signal segment, the spectral analysis results of the overlapping region are weighted and fused according to the fusion weights adjusted in S1012 to obtain the frequency domain noise distribution information of the overlapping region. Specifically, for the overlapping region of two adjacent signal segments, assuming the spectral analysis result of the overlapping region in the preceding signal segment is denoted as A, and the spectral analysis result in the following signal segment is denoted as B, the final frequency domain noise distribution information of the overlapping region is: C = w × A + (1 - w) × B; where w is the fusion weight determined in S1012. When w is 0.8, the final result of the overlapping region is more inclined towards the information of the preceding signal segment; when w is 0.2, it is more inclined towards the information of the following signal segment; and when w is 0.5, the two are equally weighted averages. For non-overlapping regions, the spectral analysis result of the segment itself is directly used without fusion processing.

[0110] Through the above adjustment mechanism, this embodiment achieves adaptive configuration of overlapping region parameters: when the signal changes drastically, the overlap length is shortened and the fusion weight is reduced to avoid using "outdated" information to contaminate the current analysis, ensuring the real-time performance of the spectrum analysis; when the signal changes smoothly, the overlap length is increased and the fusion weight is improved to fully utilize historical information to smooth the analysis results, ensuring the continuity of the spectrum analysis. Those skilled in the art can select appropriate parameter ranges and adjustment methods based on the above description and specific application scenarios to achieve dynamic adjustment of the overlapping region.

[0111] After obtaining the frequency domain noise distribution information of each signal segment, the intelligent control system executes S301 to acquire the noise feature sequence of each signal segment. The noise feature sequence refers to the sequence formed by arranging the noise features (at least including frequency domain noise distribution information) of each signal segment in chronological order, denoted as N=[n1, n2, n3, ..., nk], where ni represents the noise feature of the i-th signal segment.

[0112] Next, the intelligent control system executes S302, analyzing the noise feature correlation between adjacent signal segments and determining the influence features corresponding to each signal segment based on the noise feature correlation. The influence features characterize the range and / or degree of influence of the corresponding signal segment on subsequent signal segments. In this embodiment, the noise feature correlation is quantified using cosine similarity. Cosine similarity characterizes the degree of similarity between two signal segments by measuring the directional consistency of their frequency domain noise distribution information in multidimensional space. The calculation process of cosine similarity is as follows:

[0113] The frequency domain noise distribution information of each signal segment is considered as a multi-dimensional vector, with each dimension corresponding to the power spectral density value at a frequency point. For two adjacent signal segments, their frequency domain noise distribution vectors are obtained respectively. Then, the cosine of the angle between these two vectors is calculated, which is the cosine similarity between the two segments. The cosine similarity value ranges from -1 to 1. The closer the value is to 1, the more consistent the directions of the two vectors are, meaning the noise characteristics of the two signal segments are more similar; the closer the value is to 0, the more orthogonal the two vectors are, meaning the noise characteristics of the two signal segments are significantly different; the closer the value is to -1, the opposite the directions of the two vectors are, which is relatively rare in practical signal processing. In actual calculations, the cosine similarity value is always between 0 and 1 (because the power spectral density value is non-negative). Therefore, the cosine similarity can be directly used as a quantitative indicator of the correlation of noise characteristics, denoted as ρ. The larger the ρ value, the greater the influence of the previous signal segment on the subsequent signal segment; the smaller the ρ value, the smaller the influence.

[0114] Next, the intelligent control system determines the impact characteristics of each signal segment based on the noise feature correlation analysis results. These impact characteristics include the impact range and the degree of impact. The impact range refers to the number of subsequent signal segments that the noise characteristics of the current signal segment can affect. In this embodiment, the impact range is determined based on a preset correlation threshold T_range (this threshold can be obtained by calibrating the system before actual operation. The calibration method is as follows: in typical application scenarios, a large amount of cosine similarity data between adjacent signal segments is collected, and their distribution is statistically analyzed. A certain quantile of the statistical results is taken as the threshold, for example, the 80th quantile is taken as the strong correlation threshold T_high, and the 20th quantile is taken as the weak correlation threshold T_low. T_range can be set between T_low and T_high as needed, for example, T_range=0.6. This threshold indicates that when the cosine similarity is greater than or equal to T_range, a significant impact is considered to exist; when the cosine similarity is less than T_range, the impact is considered negligible). The specific steps are as follows:

[0115] For the i-th signal segment, calculate its cosine similarity with the (i+1), (i+2), (i+3), ... signal segments, denoted as ρ(i, i+1), ρ(i, i+2), ρ(i, i+3), ... respectively. Starting from the first subsequent signal segment, determine whether each cosine similarity is greater than or equal to the threshold T_range:

[0116] If ρ(i, i+1) ≥ T_range, then the i-th signal segment is considered to have an influence on the (i+1)-th signal segment, and the process continues to evaluate ρ(i, i+2); if ρ(i, i+2) ≥ T_range, then the i-th segment is considered to have an influence on the (i+2)-th segment, and the process continues to evaluate ρ(i, i+3); and so on, until the first signal segment with a cosine similarity less than T_range is encountered. The range of influence is the number of consecutive segments from the (i+1)-th signal segment to the last signal segment that satisfies ρ ≥ T_range. For example, if ρ(i, i+1) = 0.8, ρ(i, i+2) = 0.7, ρ(i, i+3) = 0.5, and the threshold T_range = 0.6, then ρ(i, i+1) and ρ(i, i+2) are both greater than or equal to the threshold, while ρ(i, i+3) is less than the threshold. Therefore, the influence range of the i-th signal segment is 2, meaning it affects the first and second subsequent signal segments, but not the third and subsequent signal segments. If ρ(i, i+1) is already less than T_range, then the influence range is 0, indicating that the current signal segment has no significant impact on subsequent segments.

[0117] The degree of influence refers to the magnitude of the impact of the current signal segment on subsequent signal segments. In this embodiment, the degree of influence is directly quantified using the numerical value of cosine similarity, while also considering the attenuation characteristic with increasing distance. The specific determination method is as follows:

[0118] For directly adjacent signal segments: the influence of the current signal segment on the next adjacent signal segment is directly represented by the cosine similarity ρ(i, i+1) between the two. The larger ρ(i, i+1) is, the greater the influence.

[0119] For non-adjacent signal segments separated by multiple signal segments: the influence of the current signal segment on a segment k segments apart (i.e., the (i+k)th signal segment) needs to consider two factors: 1. The cosine similarity ρ(i, i+k) between the current signal segment and the (i+k)th signal segment; 2. The influence transmitted through intermediate signal segments (i+1 to (i+k-1)th signal segments) may be attenuated. In practical implementation, one of the following two methods can be used to determine the degree of influence: Method 1: Directly use ρ(i, i+k) as the degree of influence of the i-th segment on the (i+k)-th segment. This method is simple and direct, suitable for scenarios where the influence is considered to be mainly transmitted through direct coupling rather than indirect transmission. Method 2: Represent the degree of influence as the cumulative correlation between adjacent segments on the path from i to i+k. For example, the product of ρ(i, i+1), ρ(i+1, i+2), ..., ρ(i+k-1, i+k) can be calculated, or the minimum of these values ​​can be taken as the degree of influence of the i-th segment on the (i+k)-th segment. This approach is suitable for scenarios where the impact is considered to need to be passed down through levels. Regardless of the method used, the value of the impact level ranges from 0 to 1, with a larger value indicating a stronger impact and a smaller value indicating a weaker impact.

[0120] After obtaining the influence characteristics of each signal segment, step S1021 is executed. Based on the noise characteristics of the current time segment and the corresponding influence characteristics, the initial noise evolution trend of the signal segments at subsequent time moments is predicted, and the prediction results are corrected to obtain the final noise evolution trend. Before executing S1021, the concept of an event unit needs to be explained. After the overlapping segmentation and weighted fusion processing in steps S1011 to S1013, the phase-locked loop output signal is divided into a series of continuous time units. Each time unit corresponds to a unique time interval, and adjacent time units are seamlessly connected without overlap or gaps. The specific division method is as follows:

[0121] For the overlapping region between two adjacent signal segments, this overlapping region is treated as an independent time unit, and its noise characteristics are directly determined by the weighted fusion process in step S1013. For the non-overlapping part of each signal segment that does not belong to the overlapping region, this part is also treated as an independent time unit, and its noise characteristics are directly obtained from the spectral analysis results of that segment in that region. For example, for segments i and i+1, three consecutive time units will be generated: Unit A: the non-overlapping region of segment i, with noise characteristics n_A; Unit B: the overlapping region of segment i and segment i+1, with noise characteristics n_B (obtained by weighted fusion); Unit C: the non-overlapping region of segment i+1, with noise characteristics n_C. These time units are arranged sequentially on the time axis to form a complete and continuous noise characteristic time series, denoted as N = [n1, n2, n3, ..., nk], where each n corresponds to an independent time unit, and adjacent units are continuous and seamless in time. The current time unit and subsequent time units mentioned in this step refer to the various units in the continuous time series.

[0122] Furthermore, it needs to be explained that in S302, the intelligent control system establishes a statistical model for influence transmission by analyzing historical data. The specific monitoring process is as follows: The intelligent control system records the noise characteristic sequences of each time unit over a past period, as well as the cosine similarity between them. For each time unit, the correlation distribution between it and subsequent time units is statistically analyzed. Through the statistical analysis of a large number of samples, a general rule can be derived: the noise characteristics of the current unit are usually strongly correlated with the 1st and 2nd subsequent units, the correlation with the 3rd unit weakens, and it is basically unrelated to the 4th and subsequent units. Based on this, the influence range K is determined (e.g., K=2 or K=3). For each position within the influence range (subsequent 1st unit, 2nd unit, ... Kth unit), the average or median of the cosine similarity between the current unit and the unit at that position is calculated from a large number of samples, which is taken as the degree of influence at that position. For example, the average influence of the current time unit on the first subsequent time unit is α1 = 0.8; the average influence of the current time unit on the second subsequent time unit is α2 = 0.6; and the average influence of the current time unit on the third subsequent time unit is α3 = 0.3. These influence levels α1, α2, and α3 are model parameters stored in the intelligent control system. They do not change in real time with each time unit, but are used as statistical regularities over a long period (and can be updated periodically). When making predictions in step S1021, the intelligent control system directly calls these pre-statistically calculated influence level model parameters, instead of calculating the cosine similarity between future units and the current unit in real time.

[0123] Specifically, the initial noise evolution trend refers to the result of predicting the noise characteristics of multiple consecutive future signal segments based on the noise characteristics and influence characteristics of the current signal segment. Its final form is a time series, which contains the predicted noise characteristics of the first time unit after the current time unit, the second time unit, the third time unit, and so on. This prediction sequence corresponds one-to-one with the optimal filter bandwidth sequence that needs to be matched subsequently: the first value in the prediction sequence corresponds to the bandwidth matching basis of the first subsequent time unit, the second value corresponds to the bandwidth matching basis of the second subsequent time unit, and so on. The specific method for generating the initial noise evolution trend is as follows:

[0124] First, determine the prediction range, the length of which is determined by the influence range of the current time unit. Assuming the influence range of the current time unit is K, meaning the noise characteristics of the current time unit can affect the subsequent K time units, then the prediction range is also K, meaning only the noise characteristics of the first to the Kth subsequent time units are predicted. For subsequent time units that exceed the influence range, since the current unit has no significant impact on them, no prediction is needed; their noise characteristics will be predicted by the new current time unit in subsequent steps.

[0125] Then, the prediction benchmark is determined. The prediction benchmark is the noise characteristics of the current time unit, that is, the frequency domain noise distribution information.

[0126] Next, predicted values ​​are generated sequentially based on the degree of influence: For each subsequent time unit within the prediction range, the intelligent control system calls the pre-calculated parameters of the influence degree model to generate the predicted noise characteristics for that subsequent time unit. The specific generation method is as follows:

[0127] For the next time unit immediately following the current time unit, the intelligent control system uses the influence level α1 (i.e., the average influence of the current unit on the first subsequent time unit). α1 is a value between 0 and 1, representing, statistically speaking, the proportion by which the noise characteristics of the current time unit will carry over to the next time unit. The noise characteristics of the current time unit (i.e., the power spectral density value at each frequency point) are reduced by α1, and the result is the predicted noise characteristics for the next time unit immediately following the current time unit. That is, for the next time unit immediately following the current time unit, the predicted value = noise characteristics of the current unit × α1.

[0128] For subsequent time units separated by two time units, the intelligent control system invokes an influence level α2 (i.e., the average influence of the current unit on the second subsequent time unit). α2 is a value between 0 and 1, representing, statistically speaking, the proportion by which the noise characteristics of the current time unit will carry over to the next time unit. Multiplying the noise characteristics of the current time unit by α2 yields the predicted noise characteristics for that second subsequent time unit. This process continues, with the intelligent control system invoking an influence level α2... j Multiply the noise characteristics of the current time unit by α j This yields the predicted noise features for the j-th subsequent time unit. The generated predicted values ​​for each time unit are then arranged chronologically to obtain the initial noise evolution trend sequence. This sequence contains the predicted noise features from the first time unit after the current time unit up to the k-th time unit.

[0129] After generating the initial noise evolution trend sequence, the prediction results need to be corrected to further improve prediction accuracy. The correction is based on the noise change magnitude between the current time unit and the previous time unit.

[0130] First, the noise characteristics of the current time unit are compared with those of the previous time unit, and the noise variation amplitude is calculated. The calculation method for the noise variation amplitude is the same as that for the signal change rate in step S1012, i.e., it is obtained by calculating the degree of difference in the frequency domain noise distribution information between the two time units. Specifically, for each frequency point, the absolute value of the difference between the power spectral density value of the current time unit and the previous time unit at that frequency point is calculated. Then, these absolute values ​​at all frequency points are summed to obtain a total value. This total value is the noise variation amplitude. The larger this value, the more drastic the signal change; the smaller the value, the smoother the signal change.

[0131] The correction weight is a value between 0 and 1, used to control the proportion of the correction term in the final prediction. The magnitude of the correction weight is positively correlated with the noise change amplitude: the larger the noise change amplitude, the more drastic the current signal change, and the initial prediction based on historical statistical patterns may be lagging, thus requiring a larger weight for the correction term. To quantify and determine the correction weight, this application pre-sets and stores two thresholds: a low change amplitude threshold and a high change amplitude threshold. These two thresholds can be calibrated by the intelligent control system before actual operation. The calibration method is as follows: In a typical application scenario, a large amount of noise change amplitude data between time units is collected, and its distribution range is statistically analyzed. The 20th percentile of the statistical range is taken as the low change amplitude threshold, and the 80th percentile is taken as the high change amplitude threshold. The specific rules for determining the correction weight are as follows:

[0132] When the noise change amplitude is less than or equal to the low change amplitude threshold, the signal change is considered to be gradual and the reliability of the initial prediction is relatively high. In this case, the correction weight is set to 0, indicating that no correction is performed and the initial prediction value is used completely.

[0133] When the noise change amplitude is greater than or equal to the high change amplitude threshold, the signal is considered to be changing drastically, and the initial prediction may be severely lagging. In this case, the correction weight is set to 1, indicating that the correction term is fully adopted and the initial prediction value is replaced.

[0134] When the noise change amplitude is between the low change amplitude threshold and the high change amplitude threshold, the correction weight changes linearly between 0 and 1: the closer the noise change amplitude is to the low threshold, the closer the correction weight is to 0; the closer the noise change amplitude is to the high threshold, the closer the correction weight is to 1.

[0135] For example, let the low change threshold be 30 and the high change threshold be 80. If the current noise change is 50, then first calculate the relative position of this value in the range of 30 to 80: (50-30) / (80-30)=20 / 50=0.4, and the adjustment weight is 0.4.

[0136] The correction term is generated based on the current trend and is used to predict the continuation of this trend in future time units. First, the direction of change in the current time unit relative to the previous time unit is determined. The direction of change can be determined by comparing the overall noise levels of the two units: if the overall noise characteristic of the current unit is greater than that of the previous time unit, the direction of change is considered to be increasing; if the overall noise characteristic of the current unit is less than that of the previous time unit, the direction of change is considered to be decreasing. In the frequency domain, the direction of change can be reflected as a general increase or decrease in the power spectral density at each frequency point. Second, the magnitude of the noise change is recorded, denoted as Δ. This Δ is the noise change magnitude value calculated earlier. Then, it is assumed that this trend will continue in subsequent time units, but its influence will gradually weaken over time. Based on this assumption, a correction term sequence is generated, which contains correction values ​​for each subsequent time unit. The rules for generating the correction term sequence are as follows:

[0137] For the next time unit immediately following the current time unit, the correction term is directly related to the magnitude of the current noise change and takes the value Δ. This means that it is assumed that the current trend of change will continue completely in the next unit.

[0138] For subsequent units separated by two time units, the correction term is half of Δ, i.e., Δ / 2. This indicates that the effect of the change is reduced to half in the second unit.

[0139] For subsequent units three time units later, the correction term is one-quarter of Δ, i.e., Δ / 4. This means that the effect of the change is halved again in the third unit.

[0140] Similarly, for subsequent units spaced j time units apart, the correction term is Δ divided by 2 to the power of (j-1). That is, the correction term decays exponentially with increasing distance.

[0141] It should be noted that the length of the correction term sequence should be consistent with the length of the initial predicted noise evolution trend sequence; that is, correction terms should only be generated for subsequent units within the influence range K. For units outside the influence range, no correction terms need to be generated.

[0142] Finally, the initial predicted noise evolution trend sequences are weighted and fused to obtain the final predicted noise evolution trend sequence for each subsequent time unit. For each subsequent time unit, its final predicted noise feature is obtained by a weighted combination of two parts: one part is the initial predicted value, i.e., the value predicted based on historical statistical patterns (degree of influence); the other part is the correction term, i.e., the corrected value generated based on the current trend. The specific method of weighted combination is as follows:

[0143] The final predicted value = (1 - correction weight of the current time unit × position attenuation coefficient of the current time unit) × initial predicted value + (correction weight of the current time unit × position attenuation coefficient of the current time unit) × correction term. Here, the correction weight of the current time unit is the value between 0 and 1 determined earlier based on the noise variation amplitude. This weight remains constant throughout the entire correction term sequence, representing the overall strength of this correction. The position attenuation coefficient of the current time unit reflects the attenuation of the correction term's influence with increasing distance. For the immediately following time unit, the position attenuation coefficient is 1; for subsequent units two units apart, it is 1 / 2; for subsequent units three units apart, it is 1 / 4; and so on. The position attenuation coefficient follows the same attenuation pattern as when the correction term is generated. The initial predicted value is the value determined earlier based on the influence level α. j The generated value. The correction term is the previously generated Δ divided by 2 raised to the power of (j-1).

[0144] After obtaining the final noise evolution trend, step S1022 is executed to dynamically match the optimal filter bandwidth sequence for the current and subsequent time units based on the noise evolution trend. Unlike the single-point bandwidth matching based on the noise characteristics of the current single time unit in S102, this step matches a bandwidth sequence, denoted as: B = [b_current, b_next1, b_next2, b_next3, ...]. Here, b_current represents the optimal filter bandwidth corresponding to the current time unit, b_next1 represents the optimal filter bandwidth corresponding to the first subsequent time unit, b_next2 represents the optimal filter bandwidth corresponding to the second subsequent time unit, and so on. This bandwidth sequence corresponds one-to-one with the noise evolution trend sequence finally obtained in step S1021. Each predicted value in the noise evolution trend sequence corresponds to a predicted noise characteristic for a future time unit; each bandwidth value in the bandwidth sequence is the optimal filter bandwidth for that future time unit. In this way, this step not only determines the filter bandwidth that should be used at the current moment, but also pre-plans the bandwidth that should be used for multiple consecutive time units in the future, providing a complete bandwidth configuration scheme for subsequent time-division adjustments.

[0145] The bandwidth sequence matching is also based on the noise-bandwidth mapping table mentioned above, but the input is the entire noise evolution trend sequence, not just a single noise feature at the current moment. The specific matching process is as follows:

[0146] The first step is to determine the matching range, the length of which is determined by the influence range K determined in step S302. Since the length of the noise evolution trend sequence is K (i.e., the noise characteristics of the subsequent 1st to Kth time units are predicted), the length of the bandwidth sequence is also K, that is, matching the optimal bandwidth corresponding to each of the subsequent 1st to Kth time units. For subsequent time units that are outside the influence range, since the current unit has no significant impact on them, there is no need to pre-match the bandwidth at the current time; their bandwidth will be matched by the new current time unit in subsequent steps. The second step is to match by looking up the table sequentially: for each predicted value in the noise evolution trend sequence, the corresponding optimal bandwidth is sequentially looked up in the noise-bandwidth mapping table:

[0147] For the noise feature of the current time unit (i.e., the noise feature actually detected, denoted as n_current), look up the table to obtain b_current. For the predicted noise feature n_next1 of the first subsequent time unit, look up the table to obtain b_next1. For the predicted noise feature n_next2 of the second subsequent time unit, look up the table to obtain b_next2. And so on, until the predicted noise feature n_nextK of the Kth subsequent time unit, where look up the table to obtain b_nextK.

[0148] The third step is to arrange the bandwidth values ​​obtained from the table above in chronological order to obtain the complete optimal filter bandwidth sequence: B=[b_current, b_next1, b_next2, ..., b_nextK]. This bandwidth sequence has the same length as the noise evolution trend sequence, and the elements correspond one-to-one.

[0149] Next, the intelligent control system executes S103 to analyze the direction and intensity of influence transmission between time units. The direction of influence transmission refers to the flow of influence between time units, and can be divided into three types:

[0150] Forward Influence: The noise characteristics of the current time unit affect subsequent time units. This is the most common signal processing scenario, conforming to the law of temporal causality, i.e., the past influences the future. Backward Influence: The noise characteristics of subsequent time units affect the processing of the current time unit. This may occur in systems with feedback loops, such as when the output signal of a phase-locked loop affects itself after passing through the feedback loop; or when the signal has a repetitive structure, such as subsequent periods of a periodic signal potentially containing correction information for the previous period. Bidirectional Influence: Forward and backward time units influence each other. That is, there is a mutual coupling relationship between the current unit and the subsequent unit.

[0151] The direction of propagation is influenced by analyzing the time symmetry of the correlation between noise characteristics. The specific analysis steps are as follows:

[0152] The first step is to calculate the correlation coefficient sequence between adjacent time units. For a continuous time unit sequence T = [t1, t2, t3, ..., tn], the correlation coefficient between each pair of adjacent time units is calculated sequentially, resulting in a correlation coefficient sequence R = [r12, r23, r34, ..., r(n-1)n]. Here, r12 represents the correlation coefficient between t1 and t2, r23 represents the correlation coefficient between t2 and t3, and so on. The method for calculating the correlation coefficient is the same as the cosine similarity used in step S302 when determining the influencing features.

[0153] The second step is to calculate the inverse correlation coefficient sequence. For the same time unit sequence, the correlation coefficients between adjacent time units are calculated from back to front, resulting in the inverse correlation coefficient sequence R_rev = [r21, r32, r43, ..., rn(n-1)]. Here, r21 represents the correlation coefficient between t2 and t1 (i.e., the correlation from back to front), r32 represents the correlation coefficient between t3 and t2, and so on.

[0154] The third step is to compare the positive and negative correlation coefficients. For each pair of adjacent time units, compare the positive correlation coefficient rij with the negative correlation coefficient rji: if rij is significantly greater than rji (e.g., rij > rji + preset threshold), then the forward influence is considered dominant. This indicates that the influence of t1 on t2 is greater than the influence of t2 on t1. If rji is significantly greater than rij (e.g., rji > rij + preset threshold), then the backward influence is considered dominant. This indicates that the influence of t2 on t1 is greater than the influence of t1 on t2. If the difference between rij and rji is within the preset threshold range (e.g., |rij - rji| ≤ preset threshold), then a bidirectional influence is considered to exist, meaning that the influence between the preceding and following time units is roughly equal. The preset threshold can be obtained through calibration by the intelligent control system, for example, 0.1 or 0.2, indicating that when the difference in correlation coefficients exceeds this value, it is determined that one direction is dominant.

[0155] The fourth step is to statistically determine the overall influence propagation direction. By statistically analyzing the results of comparing a large number of adjacent time unit pairs, the overall influence propagation characteristics of the phase-locked loop (PLL) signal can be determined. For example, if more than 80% of adjacent time unit pairs show a predominantly forward influence, then the PLL signal can be considered to have a predominantly forward influence overall. This statistical result will serve as the basis for determining the timing of adjustments in subsequent steps. During real-time operation, the intelligent control system can continuously update this statistical result to adapt to changes in signal characteristics.

[0156] The intelligent control system determines the optimal bandwidth adjustment timing for the current and subsequent moments based on the direction of influence transmission. Determining the adjustment timing requires a quantifiable indicator of "how many time units to delay." The adjustment timing is represented by a delay value D, where D is a non-negative integer, indicating how many time units should be delayed before bandwidth adjustment. D=0 indicates immediate adjustment, D=1 indicates adjustment after a 1-time-unit delay, D=2 indicates adjustment after a 2-time-unit delay, and so on.

[0157] When the phase-locked loop (PLL) signal is predominantly influenced by the forward direction, it means that the information in the current time unit has strong guiding significance for the future and will not be significantly corrected by future information. In this case, an immediate adjustment strategy should be adopted, i.e., delay D=0. This allows for an immediate response after detecting the noise characteristics of the current unit, improving the system's real-time performance.

[0158] When the overall PLL signal is dominated by backward influence, it means that the information in the current time unit may be corrected by the information in subsequent units, and immediate adjustment may lead to misadjustment. In this case, a delayed adjustment strategy should be adopted, and the specific value of the delay needs to be determined based on the strength of the backward influence. The method for determining the delay amount D is as follows: First, calculate the average strength of the backward influence. The strength of the backward influence can be quantified by the average value of the backward correlation coefficient rji, denoted as β_avg. Then, determine the delay amount based on the magnitude of β_avg: if β_avg ≥ 0.8, it indicates that the backward influence is very strong, and D = 2 (delay for 2 time units); if 0.6 ≤ β_avg < 0.8, then D = 1 (delay for 1 time unit); if β_avg < 0.6, then D = 0. For example, when β_avg = 0.7, the delay amount D = 1, indicating that it is necessary to wait for confirmation of the information in the next time unit before performing bandwidth adjustment for the current unit.

[0159] When the phase-locked loop (PLL) signal exhibits bidirectional influence, the delay amount needs to be determined based on the ratio of the forward influence strength to the backward influence strength. Let the forward influence strength be α (quantized using the average value of the positive correlation coefficient rij) and the backward influence strength be β (quantized using the average value of the negative correlation coefficient rji), and calculate the ratio γ = α / β. The delay amount D is determined as follows: if γ ≥ 2, it indicates that the forward influence is dominant, and D = 0 (immediate adjustment); if 1 < γ < 2, D = 0 (still forward-dominant, immediate adjustment); if 0.5 ≤ γ ≤ 1, it indicates that the forward and backward influences are roughly equal, and D = 1 (delay by 1 time unit, awaiting confirmation); if γ < 0.5, it indicates that the backward influence is dominant, and it is treated as if the backward influence is dominant.

[0160] After determining the delay amount D, when performing bandwidth adjustment, the intelligent control system does not immediately adjust the bandwidth b_current that matches the currently detected noise feature. Instead, it waits for D time units before performing the adjustment. During the waiting period, the intelligent control system continues to monitor information from subsequent units to verify or correct the adjustment decision. For example, if D=1, the bandwidth adjustment corresponding to the noise feature detected in the current time unit will be performed after the information of the next time unit is obtained; if D=2, it waits for two time units.

[0161] Next, the intelligent control system determines the execution step size for bandwidth adjustment based on the intensity of the influence transmission. Execution compensation refers to the amount of change in bandwidth adjustment for each step. The intensity of the influence transmission is the degree of influence α determined in step S302. j For the current time unit, its influence on subsequent units is α1, α2, ..., α... KHere, α1 represents the degree of influence on the immediately following time unit, α2 represents the degree of influence on the unit two time units apart, and so on. When determining the execution step size, the degree of influence α1 of the current unit on the immediately following unit is mainly considered, because this is the most direct and significant influence.

[0162] The execution step size ΔB_step represents the change in bandwidth during each adjustment, and its value is preset between the minimum step size ΔB_min and the maximum step size ΔB_max. The minimum and maximum step sizes can be determined based on the filter's adjustable accuracy and the system's stability requirements. For example, if the filter's adjustable bandwidth range is 1MHz to 100MHz and the adjustable accuracy is 0.1MHz, then ΔB_min = 0.1MHz and ΔB_max = 1MHz can be set. The specific value of the execution step size is determined by the magnitude of α1; the principle is: the larger α1 is, the smaller the execution step size; the smaller α1 is, the larger the execution step size. The specific calculation method is as follows:

[0163] First, the intelligent control system presets two thresholds: a low-influence threshold α_low and a high-influence threshold α_high. These two thresholds can be obtained through the calibration of the intelligent control system, for example, α_low = 0.3 and α_high = 0.8. Then, the execution step size is determined based on the value of α1.

[0164] When α1≥α_high, it indicates that the influence intensity is very large, and the minimum step size should be used for careful adjustment. ΔB_step=ΔB_min should be taken.

[0165] When α1≤α_low, it means that the influence intensity is very small, and the maximum step size should be used for rapid adjustment, and ΔB_step=ΔB_max should be taken;

[0166] When α_low < α1 < α_high, the execution step size varies linearly between ΔB_min and ΔB_max. First, the relative position of α1 within the interval from α_low to α_high is calculated: p = (α1 - α_low) / (α_high - α_low). Since a larger α1 results in a smaller step size, the formula for calculating the execution step size is: ΔB_step = ΔB_max - (ΔB_max - ΔB_min) × p. After determining the execution step size, the intelligent control system must ensure that each adjustment in bandwidth adjustment does not exceed ΔB_step. If a larger adjustment is required, it needs to be performed gradually in multiple steps to reach the desired value.

[0167] Finally, the intelligent control system adjusts the actual bandwidth of the filters in the phase-locked loop according to the optimal filter bandwidth sequence and execution step size, at the optimal bandwidth adjustment timing determined in S1032. The adjustment process can be implemented in steps:

[0168] Before making any adjustments, the intelligent control system has determined the bandwidth sequence B = [b_current, b_next1, b_next2, ..., b_nextK], the current actual bandwidth B_actual, the execution step size ΔB_step, and the delay D. For the bandwidth b_current at the current moment, the adjustment process is as follows:

[0169] The first step is to wait for a delay: if D > 0, the intelligent control system waits for D time units, during which time it continues to monitor the information of subsequent units, but does not perform any adjustments.

[0170] The second step is to calculate the total adjustment: Total adjustment ΔB_total = b_current - B_actual. ΔB_total may be positive (requiring increased bandwidth), negative (requiring decreased bandwidth), or zero (no adjustment required).

[0171] The third step is to determine the number of adjustments: The number of adjustments N = ceil(|ΔB_total| / ΔB_step), which is the absolute value of the total adjustment divided by the step size and then rounded up. For example, if |ΔB_total| = 2.5MHz and ΔB_step = 0.46MHz, then N = ceil(2.5 / 0.46) = ceil(5.43) = 6 times.

[0172] The fourth step is to determine the actual magnitude of each adjustment: the magnitude of the first N-1 adjustments is ΔB_step, and the magnitude of the last adjustment is the margin. If ΔB_total is positive, then increase ΔB_step each time, and increase ΔB_total-(N-1)×ΔB_step for the last time; if ΔB_total is negative, then decrease ΔB_step each time, and decrease |ΔB_total|-(N-1)×ΔB_step for the last time.

[0173] Step 5: Distribute the adjustments: Perform the above adjustments sequentially at predetermined time intervals (e.g., once per time unit). After N adjustments, check if the actual bandwidth has reached b_current. If not (possibly due to hardware response errors), a fine-tuning correction can be performed.

[0174] For the bandwidths b_next1, b_next2, etc., corresponding to subsequent time points, the adjustment process is similar, but it needs to be combined with the execution step size corresponding to that time point (this step size is determined based on the new influence intensity at that time point, not the step size at the current time point). The specific steps are as follows: First, reach the corresponding time point. When the system time advances to the start time of the time unit corresponding to b_next1, prepare to execute the adjustment of b_next1. Second, obtain the execution step size at that time point. At this time, the intelligent control system has recalculated the execution step size ΔB_step1 corresponding to that time point based on the latest influence intensity at that time point. Third, calculate the current actual bandwidth. The actual bandwidth at this time may be the value after the previous adjustment, denoted as B_actual1. Fourth, calculate the total adjustment amount. ΔB_total1 = b_next1 - B_actual1. Fifth, adjust to b_next1 step by step according to the same method as the current time point.

[0175] This application also discloses an intelligent dynamic response phase-locked loop filter system. (Refer to...) Figure 2 ,include:

[0176] The signal monitoring module 201 is used to perform segmented collaborative processing on the phase-locked loop output signal to detect the noise characteristics of the output signal in real time. The noise characteristics include at least the frequency domain noise distribution information of the output signal.

[0177] The bandwidth matching module 202 is used to dynamically match the optimal filter bandwidth under the current noise environment based on the noise characteristics.

[0178] The bandwidth adjustment module 203 is used to adjust the actual bandwidth of the filter in the phase-locked loop according to the optimal filter bandwidth.

[0179] This application also discloses an intelligent dynamic response phase-locked loop filter device, which includes a memory and a processor. The memory stores a computer program that can be loaded by the processor and executed as described above for the intelligent dynamic response phase-locked loop filter method.

[0180] This application also discloses a computer-readable storage medium that stores a computer program capable of being loaded by a processor and executed as described above for a smart dynamic response phase-locked loop filtering method. The computer-readable storage medium includes, for example, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

Claims

1. A phase-locked loop filtering method with intelligent dynamic response, characterized in that, include: The output signal of the phase-locked loop is segmented on the time axis, and there is an overlapping area between adjacent signal segments, forming multiple signal segments that are continuously arranged in time. The signal change rate of the output signal is detected, and the length of the overlapping region and / or the fusion weight of the overlapping region are dynamically adjusted according to the signal change rate. Spectral analysis is performed on each signal segment in parallel to obtain the frequency domain noise distribution information of each signal segment; according to the adjusted fusion weight, the spectral analysis results of the overlapping region are weighted and fused to obtain the frequency domain noise distribution information of the overlapping region, so as to detect the noise characteristics of the output signal in real time, wherein the noise characteristics include at least the frequency domain noise distribution information of the output signal; Based on the noise characteristics, the optimal filter bandwidth under the current noise environment is dynamically matched through a preset mapping table that maps noise characteristics to optimal bandwidth; the mapping table records the optimal bandwidth value that should be used under different noise conditions. Adjust the actual bandwidth of the filter in the phase-locked loop according to the optimal filter bandwidth.

2. The intelligent dynamically responsive phase-locked loop filtering method of claim 1, wherein, The method further includes: Real-time monitoring of the ambient temperature of the phase-locked loop's environment, combined with historical temperature data, to predict temperature change trends; Based on the temperature change trend, the compensation amount of the control parameters of the voltage-controlled oscillator is determined, and the control parameters of the voltage-controlled oscillator are adjusted according to the compensation amount to compensate for the frequency drift caused by temperature changes.

3. The intelligent dynamically responsive phase-locked loop filtering method of claim 1, wherein, The method further includes: Detect whether the phase-locked loop has entered a stable locked state, which means that the phase-locked loop has established a stable frequency and phase tracking state; When the phase-locked loop enters a stable locked state, at least one high-power module in the phase-locked loop is turned off, so that the phase-locked loop enters a low-power mode; the high-power module refers to the circuit unit in the phase-locked loop whose power consumption is higher than a preset value in normal operation mode. In the low-power mode, the phase error of the phase-locked loop is monitored; When the phase error is detected to exceed the preset threshold, the high-power module that was shut down is restarted.

4. The intelligent dynamically responsive phase-locked loop filtering method of claim 1, wherein, The method further includes: Obtain the noise feature sequence of each signal segment; By calculating the cosine similarity of the frequency domain noise distribution information of adjacent signal segments, the correlation of noise features between adjacent signal segments is analyzed; based on the comparison result of the cosine similarity with a preset correlation threshold, the influence range corresponding to each signal segment is determined; based on the magnitude of the cosine similarity, the degree of influence of each signal segment on subsequent positions is determined; the influence range refers to the number of subsequent signal segments that the noise features of the current signal segment can affect; the degree of influence refers to the magnitude of the influence of the current signal segment on subsequent signal segments. Based on the noise characteristics, and using a preset mapping table between noise characteristics and optimal bandwidth, the optimal filter bandwidth under the current noise environment is dynamically matched, including: Based on the noise characteristics of the current signal segment and the degree of influence of the current signal segment on subsequent positions, the initial predicted noise characteristics of the signal segments at subsequent positions are generated, thereby predicting the noise evolution trend of the signal segments at each position at subsequent time points. The noise characteristics of the current signal segment are compared with the noise characteristics of the previous signal segment to obtain the noise change amplitude; the correction weight is determined based on the comparison results of the noise change amplitude with the preset low change amplitude threshold and high change amplitude threshold; the noise evolution trend is corrected according to the correction weight to obtain the final noise evolution trend, wherein the larger the noise change amplitude, the larger the correction weight. Based on the noise evolution trend, the optimal filter bandwidth sequence for the current and subsequent times is dynamically matched through a preset mapping table between noise characteristics and optimal bandwidth. The step of adjusting the actual bandwidth of the filter in the phase-locked loop according to the optimal filter bandwidth includes: Based on the optimal filter bandwidth sequence, adjust the actual bandwidth of the filter in the phase-locked loop at the current and subsequent times.

5. The intelligent dynamically responsive phase-locked loop filtering method of claim 4, wherein, The step of adjusting the actual bandwidth of the filter in the phase-locked loop at the current and subsequent times according to the optimal filter bandwidth sequence includes: Analyze the direction and intensity of influence transmission between different signal segments; Based on the direction of the influence propagation, determine the optimal bandwidth adjustment timing for the current and subsequent moments; The execution step size for bandwidth adjustment is determined based on the intensity of the influence transmission, wherein the greater the intensity of the influence transmission, the smaller the execution step size; According to the optimal filter bandwidth sequence and execution step size, at the optimal bandwidth adjustment time, the actual bandwidth of the filter in the phase-locked loop is adjusted.

6. The intelligent dynamically responsive phase-locked loop filtering method of claim 4, wherein, The step of predicting the noise evolution trend of signal segments at subsequent times based on the noise characteristics of the current signal segment and the corresponding influence characteristics includes: Based on the noise characteristics of the current time segment and the corresponding impact characteristics, predict the noise evolution trend of the signal segments in subsequent time segments; The noise characteristics of the current signal segment are compared with the noise characteristics of the previous signal segment to obtain the noise variation amplitude; The final noise evolution trend is obtained by correcting the noise evolution trend based on the noise change amplitude, wherein the larger the noise change amplitude, the greater the correction weight.

7. A phase-locked loop filter system with intelligent dynamic response, characterized in that, include, The signal monitoring module (201) is used to segment the phase-locked loop output signal on the time axis, and there is an overlapping area between adjacent signal segments to form multiple signal segments arranged continuously in time; detect the signal change rate of the output signal, and dynamically adjust the length of the overlapping area and / or the fusion weight of the overlapping area according to the signal change rate; The bandwidth matching module (202) is used to perform parallel spectrum analysis on each signal segment to obtain the frequency domain noise distribution information of each signal segment; according to the adjusted fusion weight, the spectrum analysis results of the overlapping region are weighted and fused to obtain the frequency domain noise distribution information of the overlapping region, so as to detect the noise characteristics of the output signal in real time, wherein the noise characteristics include at least the frequency domain noise distribution information of the output signal; it is also used to dynamically match the optimal filter bandwidth under the current noise environment according to the noise characteristics through a preset mapping table between noise characteristics and optimal bandwidth; the mapping table records the optimal bandwidth value that should be used under different noise conditions. The bandwidth adjustment module (203) is used to adjust the actual bandwidth of the filter in the phase-locked loop according to the optimal filter bandwidth.

8. An intelligent dynamically responsive phase-locked loop filter device, characterized by, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed as described in any one of claims 1 to 6.

Images