Frequency adjustment method, system and related device based on sip module

Abstract

The application provides a frequency adjustment method and system based on an SIP module and related equipment, and relates to the technical field of radio frequency front ends.The method comprises the following steps: calculating a temperature change rate and a humidity change rate according to environmental temperature data and environmental humidity data; inputting the temperature change rate and the humidity change rate into a preset prediction model to predict a future resonance frequency drift amount of an FBAR filter; calculating a resonance frequency adjustment amount according to the predicted resonance frequency drift amount; generating a control signal based on the resonance frequency adjustment amount according to the response characteristics and the adjustment step of an adjustable element on the FBAR filter; and adjusting the resonance frequency of the FBAR filter by using the control signal.The method of the application overcomes the inherent sensor response delay, data processing delay and frequency adjustment execution delay of the existing adjustment mechanism based on real-time feedback, realizes fast and accurate dynamic adjustment of the resonance frequency of the FBAR filter, and ensures stable radio frequency communication.

Description

Technical Field

[0001] The present invention relates to the field of radio frequency front-end technology, and in particular to a frequency adjustment method and system based on a SIP module, and related equipment. Background Art

[0002] A RF front-end module in a SIP (system-in-package) package integrates an FBAR filter (thin film bulk acoustic wave filter) and a temperature and humidity sensor, and is deployed in outdoor base stations. The ambient temperature and humidity in outdoor base stations fluctuate significantly and frequently. Specifically, during the day, sunlight can cause the module's surface temperature to rise rapidly, exceeding the ambient air temperature. At night or during seasonal changes, the ambient temperature drops rapidly, potentially dropping below freezing. This rapid alternation of high and low temperatures—for example, switching from high to low temperatures, or vice versa, in a short period of time—poses a challenge to the FBAR filter integrated into the module. FBAR filters utilize the bulk acoustic wave resonance effect of piezoelectric materials to achieve frequency selection. The resonant frequency is affected by the material's properties. The material's elastic modulus and density vary with temperature, causing changes in the propagation speed of acoustic waves and, in turn, causing the resonant frequency to drift. Increasing temperature causes the frequency to shift downward, while decreasing temperature causes the frequency to shift upward. In environments with this rapid alternation, the temperature change rate can reach several degrees Celsius per minute, accelerating the frequency drift of the FBAR filter. The module's integrated temperature and humidity sensor monitors the ambient temperature and humidity in real time. Existing dynamic adjustment methods typically rely on the current temperature value output by a temperature sensor. By looking up a pre-calibrated temperature-frequency compensation curve or table, the required frequency adjustment is determined. The frequency is then adjusted by controlling the adjustable element to change the filter parameters to correct the frequency. However, this adjustment mechanism based on real-time feedback is subject to response chain delays. First, it takes time for the temperature and humidity sensor to sense environmental changes and output stable data. Second, the data processing unit within the module takes time to read the sensor data, perform table lookup or calculation of compensation values, and generate control signals. Finally, the adjustable element used for frequency adjustment also has response speed and adjustment step size limitations when receiving the control signal and actually changing the filter parameters. When the ambient temperature changes rapidly, the frequency drift of the FBAR filter is continuous and rapid, while the adjustment process is discrete and delayed. This means that the temperature data collected at a given moment reflects the previous state, and the compensation value calculated based on this data corresponds to the previous frequency drift. However, by the time the actual adjustment is performed, the ambient temperature and FBAR frequency have already continuously changed. This delay effect is particularly pronounced when high and low temperatures fluctuate rapidly, which can result in a specific moment-to-moment error between the actual adjusted frequency and the target frequency. This error may not meet the frequency stability requirements of the communication system. For example, in narrowband communication systems, even a small frequency deviation can cause the signal to fall out of band, making demodulation impossible at the receiver. In broadband systems, frequency deviation can affect the passband characteristics, increasing insertion loss or reducing in-band flatness.

[0003] In addition, alternating high and low temperature environments are often accompanied by changes in humidity, especially when the temperature drops, condensation may occur, causing a rapid increase in local humidity. The FBAR filter itself is relatively insensitive to humidity, but the module's packaging materials, substrate materials, solder joints, wire bonds and other components may be affected by humidity. For example, the dielectric constant of hygroscopic materials changes with humidity, which may affect parasitic capacitance or dielectric loss, and thus indirectly affect the filter's key RF parameters such as resonant frequency, bandwidth, insertion loss and return loss. In high humidity or condensation environments, these indirect effects may be superimposed on the frequency drift caused by temperature, making compensation based solely on temperature insufficient. The existing technology does not fully consider the possible impact of humidity on the performance of FBAR filters, especially when temperature and humidity change jointly and at a high speed.

[0004] Therefore, in specific industrial application scenarios where high and low temperature speeds alternate and are accompanied by humidity changes, the existing dynamic adjustment method based on temperature sensor feedback cannot effectively cope with the speed frequency drift and performance changes of the FBAR filter due to its response delay, limited adjustment accuracy and failure to fully consider the impact of humidity, thus affecting the stability of RF communication. Summary of the Invention

[0005] The purpose of the present invention is to provide a frequency adjustment method, system, and related equipment based on a SIP module, which overcomes the sensor response delay, data processing delay, and frequency adjustment execution delay inherent in the existing adjustment mechanism based on real-time feedback, and realizes fast and accurate dynamic adjustment of the FBAR filter resonant frequency to ensure stable radio frequency communication.

[0006] In a first aspect, the present invention provides a frequency adjustment method based on a SIP module, which is applied to a radio frequency front-end control system, comprising the following steps:

[0007] Collecting ambient temperature data and ambient humidity data, and calculating the temperature change rate and humidity change rate based on the ambient temperature data and ambient humidity data;

[0008] Inputting the temperature change rate and the humidity change rate into a preset prediction model to predict the future resonant frequency drift of the FBAR filter; the prediction model is established based on historical temperature and humidity change data;

[0009] Calculating a resonant frequency adjustment amount based on the predicted resonant frequency drift amount;

[0010] generating a control signal based on the resonant frequency adjustment amount according to the response characteristics and adjustment step size of the adjustable element of the FBAR filter;

[0011] The control signal is used to control the adjustable element on the FBAR filter to change the filter parameters, so as to adjust the resonant frequency of the FBAR filter.

[0012] The frequency adjustment method based on the SIP module provided by the present invention monitors the rate of change of temperature and humidity in real time. Using these rate of change data, combined with historical temperature and humidity change data, a prediction model is established to predict the frequency drift of the FBAR filter in the short term in the future. Based on the predicted future frequency drift, a control signal for the FBAR filter adjustment element is calculated and generated in advance. Before or during the predicted frequency drift, the adjustment element is driven to change the filter parameters to achieve predictive frequency adjustment. This solution can initiate compensation before or during the actual frequency drift, effectively overcoming the delays in the sensor, data processing, and execution links, making the adjustment more timely and accurate. At the same time, the combined effects of temperature and humidity are taken into account, thereby improving the adjustment accuracy in high-humidity environments.

[0013] Furthermore, the step of calculating the resonant frequency adjustment amount according to the predicted resonant frequency drift includes:

[0014] Determining a maximum resonant frequency adjustment amount based on an initial resonant frequency of the FBAR filter and a preset frequency adjustment range;

[0015] Determining whether the predicted resonant frequency drift exceeds an adjustment range defined by the maximum resonant frequency adjustment amount;

[0016] If the predicted resonant frequency drift does not exceed the adjustment range, the predicted resonant frequency drift is used as the resonant frequency adjustment amount;

[0017] If the predicted resonant frequency drift exceeds the adjustment range, the resonant frequency adjustment amount is set to the maximum resonant frequency adjustment amount, and the frequency drift amount exceeding the adjustment range, the current ambient temperature data, the current ambient humidity data and the timestamp information are encapsulated as an out-of-limit drift record; the out-of-limit drift record is used to adjust the model parameters of the prediction model.

[0018] This mechanism provides a feedback loop that enables the prediction model to continuously learn and adapt over time and as data accumulates, thereby improving the overall frequency adjustment effect and the stability of the RF front-end.

[0019] Furthermore, the method further includes the following steps:

[0020] Adding the over-limit drift record to an over-limit drift database, and determining whether the number of over-limit drift records in the over-limit drift database exceeds a preset over-limit drift record number threshold;

[0021] If the number of out-of-limit drift records in the out-of-limit drift database exceeds a preset threshold number of out-of-limit drift records, all out-of-limit drift records are extracted from the out-of-limit drift database, and a temperature change sequence and a humidity change sequence are constructed based on the ambient temperature data, ambient humidity data, and timestamp information in each out-of-limit drift record;

[0022] Calculating a temperature change gradient sequence and a humidity change gradient sequence according to the temperature change sequence and the humidity change sequence, and performing time series alignment processing on the temperature change gradient sequence and the humidity change gradient sequence using a dynamic time warping algorithm to obtain aligned temperature change gradient sequence and humidity change gradient sequence;

[0023] The aligned temperature change gradient sequence and humidity change gradient sequence are used as input, and the frequency drift amount exceeding the adjustment range in each over-limit drift record is used as output, and the model parameters of the prediction model are adjusted using a gradient descent algorithm.

[0024] The prediction model can more accurately predict possible frequency drift in the future, reduce the situation where the predicted drift exceeds the maximum adjustment range, and improve the effectiveness of the overall frequency adjustment, thereby maintaining the stable operation of the RF front-end control system.

[0025] Furthermore, the steps of calculating a temperature change gradient sequence and a humidity change gradient sequence based on the temperature change sequence and the humidity change sequence, and performing time series alignment processing on the temperature change gradient sequence and the humidity change gradient sequence using a dynamic time warping algorithm to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence include:

[0026] For the temperature change sequence and the humidity change sequence, first-order difference calculation is used to obtain a temperature change gradient sequence and a humidity change gradient sequence respectively;

[0027] Based on the pre-built cost matrix, a dynamic programming algorithm is used to find the optimal regularized path;

[0028] According to the optimal regularization path, the temperature change gradient sequence and the humidity change gradient sequence are time-series aligned to obtain aligned temperature change gradient sequence and humidity change gradient sequence.

[0029] It provides high-quality input data for the subsequent use of the gradient descent algorithm to adjust the prediction model parameters, thereby improving the accuracy and robustness of the prediction model, and ultimately improving the precision and effectiveness of the FBAR filter resonant frequency adjustment.

[0030] Furthermore, according to the optimal regularization path, the temperature change gradient sequence and the humidity change gradient sequence are time-series aligned to obtain the aligned temperature change gradient sequence and humidity change gradient sequence, including the following steps:

[0031] Determine the time points that need to be aligned in the temperature change gradient sequence and the humidity change gradient sequence according to the optimal regularization path, and construct a time point mapping table; the time point mapping table records the corresponding position of each time point in the temperature change gradient sequence and the humidity change gradient sequence in the aligned sequence;

[0032] According to the time point mapping table, the temperature change gradient sequence and the humidity change gradient sequence are adjusted in time series, specifically including: if there is a time point in the time point mapping table that corresponds to multiple time points, interpolation processing is performed using a linear interpolation method; if there is a time point in the time point mapping table that has no corresponding time point, deleting the time point from the sequence;

[0033] The adjusted temperature change gradient sequence and humidity change gradient sequence are spliced ​​to obtain aligned temperature change gradient sequence and humidity change gradient sequence.

[0034] Furthermore, the step of using the aligned temperature change gradient sequence and humidity change gradient sequence as input and the frequency drift amount exceeding the adjustment range in each over-limit drift record as output, and adjusting the model parameters of the prediction model using a gradient descent algorithm includes:

[0035] Determine the type of prediction model, specifically: if the prediction model is a neural network model, use the Adam optimization algorithm as the gradient descent algorithm; if the prediction model is a support vector machine model, use the SMO algorithm as the gradient descent algorithm;

[0036] According to the ambient temperature data and ambient humidity data in the over-limit drift record, the learning rate adjustment range is determined by calculating the temperature change range and the humidity change range;

[0037] The aligned temperature change gradient sequence and humidity change gradient sequence are taken as input, and the frequency drift amount exceeding the adjustment range in each out-of-limit drift record is taken as output. The Adam optimization algorithm or the SMO algorithm is used to adaptively adjust the learning rate within the learning rate adjustment range, and the model parameters of the prediction model are adjusted based on the adjusted learning rate.

[0038] In a second aspect, the present invention provides a radio frequency front-end control system, comprising a controller, wherein the controller is configured to execute the steps in the frequency adjustment method based on the SIP module as described above.

[0039] In a third aspect, the present invention provides an adjustment device based on a SIP module, which is applied to a radio frequency front-end control system, comprising:

[0040] The acquisition module is used to collect ambient temperature data and ambient humidity data, and calculate the temperature change rate and humidity change rate based on the ambient temperature data and ambient humidity data;

[0041] A prediction module is used to input the temperature change rate and the humidity change rate into a preset prediction model to predict the future resonant frequency drift of the FBAR filter;

[0042] A calculation module, configured to calculate a resonant frequency adjustment amount based on the predicted resonant frequency drift amount;

[0043] A generating module, configured to generate a control signal based on the resonant frequency adjustment amount according to a response characteristic and an adjustment step size of an adjustable element on the FBAR filter;

[0044] The control module is used to control the adjustable element on the FBAR filter to change the filter parameters using the control signal, so as to adjust the resonant frequency of the FBAR filter.

[0045] The adjustment device based on the SIP module provided by the present invention can adjust the resonant frequency of the FBAR filter more timely and accurately even in an environment where the temperature and humidity change rapidly, thereby ensuring the performance of the communication system.

[0046] In a fourth aspect, the present invention provides an electronic device comprising a processor and a memory, wherein the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the steps in the frequency adjustment method based on the SIP module provided in the first aspect are executed.

[0047] In a fifth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the frequency adjustment method based on the SIP module provided in the first aspect are executed.

[0048] As can be seen from the above, the frequency adjustment method based on the SIP module provided by the present invention, by introducing a predictive adjustment mechanism based on the rate and trend of temperature and humidity changes, can effectively overcome the sensor response delay, data processing delay and frequency adjustment execution delay of the existing real-time feedback method in an industrial environment where high and low temperatures change rapidly and are accompanied by humidity changes. By predicting the frequency drift in the short future time and initiating the adjustment in advance, the response speed and accuracy of the dynamic adjustment of the FBAR filter are significantly improved, so that the actual frequency can track the target frequency more closely. At the same time, the prediction model comprehensively considers the combined effects of temperature and humidity, improving the compensation accuracy in high humidity or condensation environments. Ultimately, this solution can maintain the stability and reliability of radio frequency communications in highly dynamic temperature and humidity environments.

[0049] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the embodiments of the present invention. The purposes and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A flow chart of a frequency adjustment method based on a SIP module provided in an embodiment of the present invention.

[0051] Figure 2 A structural diagram of an adjustment device based on a SIP module provided in an embodiment of the present invention.

[0052] Figure 3 A schematic structural diagram of an electronic device provided by an embodiment of the present invention.

[0053] Description of labels:

[0054] 100, acquisition module; 200, prediction module; 300, calculation module; 400, generation module; 500, control module; 600, optimization module; 13, electronic device; 1301, processor; 1302, memory; 1303, communication bus. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0056] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are used only to distinguish the description and should not be understood as indicating or implying relative importance.

[0057] Reference Attachment Figure 1 The present invention provides a frequency adjustment method based on a SIP module, which is applied to a radio frequency front-end control system, comprising the steps of:

[0058] Collecting ambient temperature data and ambient humidity data, and calculating the temperature change rate and humidity change rate based on the ambient temperature data and ambient humidity data;

[0059] The temperature change rate and humidity change rate are input into a preset prediction model to predict the future resonant frequency drift of the FBAR filter; the prediction model is established based on historical temperature and humidity change data;

[0060] Calculating a resonant frequency adjustment amount based on the predicted resonant frequency drift amount;

[0061] generating a control signal based on the resonant frequency adjustment amount according to the response characteristics and adjustment step size of the adjustable element on the FBAR filter;

[0062] The control signal is used to control the adjustable elements on the FBAR filter to change the filter parameters to adjust the resonant frequency of the FBAR filter.

[0063] Ambient temperature and humidity data are acquired via a temperature and humidity sensor integrated into the SIP module. Specifically, the temperature and humidity sensor periodically measures ambient temperature and humidity and outputs the measurement results. The temperature and humidity change rates are calculated by differentially computing the continuously collected ambient temperature and humidity data, for example, calculating the temperature or humidity change per unit time. The prediction model is a mathematical model whose input is the calculated temperature and humidity change rates, and whose output is the predicted resonant frequency drift of the FBAR filter at a future point in time or time period. Before deployment, the model is trained using historically collected ambient temperature and humidity change data and the corresponding FBAR filter frequency drift data. The resonant frequency adjustment is determined based on the predicted resonant frequency drift, typically the inverse of the predicted drift to offset the predicted drift. The tunable element in the FBAR filter is, for example, a varactor diode, whose capacitance changes with an applied control voltage, thereby affecting the filter's resonant frequency. The control signal generation step converts the calculated resonant frequency adjustment into an electrical signal, such as a voltage or current, required to control the tunable element, taking into account the tunable element's response curve and minimum adjustment step size. The control signal is applied to the tunable element to change its parameters, thereby adjusting the resonant frequency of the FBAR filter.

[0064] Specifically, this method addresses the problem of delayed frequency drift adjustment caused by rapid environmental changes by predicting the future resonant frequency drift of an FBAR filter. First, the system collects current ambient temperature and humidity data and calculates the rates of change of temperature and humidity. These rates reflect the trend and speed of environmental change. Next, these rates are input into a pre-established prediction model. This prediction model is trained using a large amount of historical temperature and humidity change data and the corresponding FBAR filter frequency drift data, learning the complex relationship between environmental dynamics and frequency drift. Using this model, the system can predict the potential resonant frequency drift of the FBAR filter over a period of time. Based on the predicted future drift, the system calculates the required resonant frequency adjustment. This adjustment offsets the predicted drift, maintaining the filter frequency near the target value. The calculated resonant frequency adjustment is then converted into a specific control signal based on the characteristics of the tunable element used for frequency adjustment in the FBAR filter (for example, the voltage-capacitance curve of the varactor diode) and its minimum adjustment step size. Finally, the generated control signal is applied to the tunable element, changing its parameters and thereby adjusting the resonant frequency of the FBAR filter. This prediction-adjustment mechanism allows the system to compensate for frequency drift before or as it occurs, effectively reducing the response delay of traditional real-time feedback-based methods and improving the frequency stability of the RF front-end control system in rapidly changing environments. As a result, the resonant frequency of the FBAR filter can be adjusted more promptly and accurately, even in environments with rapidly changing temperature and humidity, ensuring the performance of the communication system.

[0065] In some embodiments, the controller in the RF front-end control system is integrated into the SIP module. A temperature and humidity sensor collects ambient temperature and humidity data 10 times per second. Every 100 milliseconds, the controller calculates the temperature and humidity change rates over the past second, for example, by calculating the difference between the current value and the value one second prior. These calculated temperature and humidity change rates are input into a prediction model based on a long short-term memory (LSTM) network, which predicts the resonant frequency drift of the FBAR filter over the next five seconds. The prediction model is trained offline using several weeks of field temperature, humidity, and FBAR frequency drift data before system deployment. Based on the predicted resonant frequency drift, the controller calculates the required frequency adjustment. For example, if a future drift of +10 kHz is predicted, the calculated adjustment is -10 kHz. A varactor diode is integrated into the FBAR filter for frequency fine-tuning. Its response characteristics (voltage-frequency curve) and minimum adjustment step size (e.g., corresponding to a 0.5 kHz frequency change) are stored in the controller's memory. Based on the calculated -10kHz adjustment, the controller consults the varactor's response characteristics to determine the required control voltage change, quantizing it by taking into account the adjustment step size. The controller then outputs the corresponding control voltage signal via a digital-to-analog converter (DAC). This control voltage signal is applied to the varactor, changing its capacitance and thereby adjusting the resonant frequency of the FBAR filter. For example, if the frequency needs to be shifted down by 10kHz, the controller calculates the corresponding control voltage and outputs it via the DAC, increasing the capacitance of the varactor and thereby lowering the resonant frequency of the FBAR filter.

[0066] In some embodiments, the step of calculating the resonant frequency adjustment amount according to the predicted resonant frequency drift includes:

[0067] Determining a maximum resonant frequency adjustment amount based on an initial resonant frequency of the FBAR filter and a preset frequency adjustment range;

[0068] Determining whether the predicted resonant frequency drift exceeds an adjustment range limited by a maximum resonant frequency adjustment amount;

[0069] If the predicted resonant frequency drift does not exceed the adjustment range, the predicted resonant frequency drift is used as the resonant frequency adjustment amount;

[0070] If the predicted resonant frequency drift exceeds the adjustment range, the resonant frequency adjustment amount is set to the maximum resonant frequency adjustment amount, and the frequency drift amount exceeding the adjustment range, the current ambient temperature data, the current ambient humidity data, and the timestamp information are encapsulated as an out-of-limit drift record; the out-of-limit drift record is used to adjust the model parameters of the prediction model.

[0071] The maximum resonant frequency adjustment is determined based on the initial resonant frequency of the FBAR filter and the preset frequency adjustment range. The initial resonant frequency is the center frequency of the filter design. The preset frequency adjustment range is the frequency offset range that the adjustable element can achieve. Therefore, the maximum resonant frequency adjustment is determined as the upper limit of the preset frequency adjustment range. A determination is made as to whether the predicted resonant frequency drift exceeds the adjustment range specified by the maximum resonant frequency adjustment. The drift output by the prediction model is compared with the maximum resonant frequency adjustment. If the predicted resonant frequency drift does not exceed the adjustment range, the predicted resonant frequency drift is used as the resonant frequency adjustment. The system attempts to compensate based on the predicted value. If the predicted resonant frequency drift exceeds the adjustment range, the resonant frequency adjustment is set to the maximum resonant frequency adjustment. The actual adjustment is limited to the capability of the adjustable element. The frequency drift that exceeds the adjustment range, the current ambient temperature data, the current ambient humidity data, and a timestamp are encapsulated into an out-of-limit drift record. This record contains information about the prediction error and the environmental conditions at the time of occurrence, which is used for subsequent model optimization.

[0072] Specifically, this technical solution addresses the problem of predicted resonant frequency drift exceeding the actual adjustment capability of the tunable element in an FBAR filter. First, the system calculates the maximum frequency adjustment the filter can make based on the initial resonant frequency of the FBAR filter and the preset frequency adjustment range of the tunable element. This maximum adjustment sets the physical upper limit for adjustment. Next, the system obtains the future resonant frequency drift output by the prediction model and compares this predicted drift with the previously determined maximum adjustment capability. If the predicted drift is less than or equal to the maximum adjustment capability, indicating that the predicted drift is within the capability of the tunable element, the system directly uses the predicted drift as the required resonant frequency adjustment and generates a corresponding control signal. If the predicted drift is greater than the maximum adjustment capability, indicating that the predicted drift exceeds the capability of the tunable element, the system does not attempt to make adjustments beyond its capability and instead limits the actual resonant frequency adjustment to the maximum resonant frequency adjustment capability. This ensures the validity of the control signal and the feasibility of the adjustment operation, avoiding system instability that could result from attempting excessive adjustments. Furthermore, to learn from this prediction error and improve the model, the system packages the portion of the predicted drift that exceeds the maximum adjustment (i.e., the frequency drift outside the adjustment range), the ambient temperature and humidity data at the time of the error, and the precise timestamp information into a data record, called an out-of-limit drift record. This record is stored for subsequent adjustment and optimization of the prediction model parameters. By collecting these out-of-limit drift records, the system can identify specific temperature and humidity fluctuations under which the prediction model is prone to large errors. This data can be used to improve the model's prediction accuracy, especially in extreme or rapidly changing temperature and humidity environments. This mechanism provides a feedback loop, allowing the prediction model to continuously learn and adapt over time and as data accumulates, thereby improving the overall frequency adjustment performance and RF front-end stability.

[0073] In some specific embodiments, assume that the initial resonant frequency of the FBAR filter is 2.5 GHz, and the preset frequency adjustment range of the tunable element is ±10 MHz. Therefore, the maximum resonant frequency adjustment is determined to be 10 MHz. At a certain moment, the prediction model predicts a future resonant frequency drift of +12 MHz based on collected ambient temperature and humidity data. The system compares the predicted +12 MHz with the maximum adjustment of 10 MHz. Because +12 MHz exceeds the ±10 MHz adjustment range, the system does not attempt to adjust the resonant frequency by +12 MHz. The actual resonant frequency adjustment is set to the maximum adjustment of +10 MHz. The system also records the frequency drift amount (12 MHz - 10 MHz = 2 MHz) that exceeds the adjustment range, the current ambient temperature (e.g., 35.2°C), the current ambient humidity (e.g., 88% RH), and a timestamp (e.g., 2023-10-27 14:30:00). This information is packaged into a drift-exceeding-limit record, such as {Drift exceeded: 2 MHz, Temperature: 35.2°C, Humidity: 88% RH, Timestamp: 2023-10-27 14:30:00}. This record is stored in memory and used for subsequent model parameter adjustments. This way, while limiting the actual adjustment amount, the system also collects valuable prediction error data, providing a foundation for improving the accuracy of the prediction model.

[0074] In some embodiments, the following steps are further included:

[0075] Adding the over-limit drift record to the over-limit drift database, and determining whether the number of over-limit drift records in the over-limit drift database exceeds a preset over-limit drift record number threshold;

[0076] If the number of out-of-limit drift records in the out-of-limit drift database exceeds a preset threshold number of out-of-limit drift records, all out-of-limit drift records are extracted from the out-of-limit drift database, and a temperature change sequence and a humidity change sequence are constructed based on the ambient temperature data, ambient humidity data, and timestamp information in each out-of-limit drift record;

[0077] According to the temperature change sequence and the humidity change sequence, the temperature change gradient sequence and the humidity change gradient sequence are calculated, and the temperature change gradient sequence and the humidity change gradient sequence are aligned using the dynamic time warping algorithm to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence;

[0078] The aligned temperature change gradient sequence and humidity change gradient sequence are taken as input, and the frequency drift amount exceeding the adjustment range in each over-limit drift record is taken as output. The gradient descent algorithm is used to adjust the model parameters of the prediction model.

[0079] When the predicted resonant frequency drift exceeds the adjustment range specified by the maximum resonant frequency adjustment, the frequency drift that exceeds the adjustment range, the current ambient temperature data, the current ambient humidity data, and a timestamp are packaged into an out-of-limit drift record. This out-of-limit drift record is added to the out-of-limit drift database, which stores these records. The system continuously determines the number of out-of-limit drift records in the out-of-limit drift database. When the number of records reaches a preset threshold, the parameter adjustment process of the prediction model is triggered. All out-of-limit drift records are extracted from the out-of-limit drift database. Each record contains ambient temperature data, ambient humidity data, and a timestamp. Based on this information, temperature and humidity change sequences are constructed to reflect the environmental change process. The temperature and humidity change sequences reflect the environmental conditions experienced when the prediction model failed. Based on the temperature and humidity change sequences, temperature and humidity change gradient sequences are calculated. These gradients reflect the rate of change of the ambient temperature and humidity. The temperature and humidity change gradient sequences are time-series aligned using a dynamic time warping algorithm. The dynamic time warping algorithm aligns the two time series by finding the optimal path, addressing any potential time offsets or rate differences in environmental changes between records. This results in aligned temperature and humidity gradient sequences. These aligned temperature and humidity gradient sequences serve as input features, and the frequency drift outside the adjustment range in each over-limit drift record serves as the corresponding output target. A gradient descent algorithm is used to adjust the model parameters based on these input and output data.

[0080] Specifically, this solution aims to address the problem of inaccurate predictions and insufficient adjustments caused by prediction models in environments subject to sudden, nonlinear temperature and humidity changes. When the predicted frequency drift exceeds the maximum adjustment capacity of the FBAR filter's adjustable element, the environmental conditions (temperature, humidity, and time) at the time of the prediction failure, as well as the actual uncompensated drift, are recorded. These records are collected in an over-limit drift database. When a sufficient number of such failure records (reaching a preset threshold) are collected, it indicates that the prediction model has systematic errors under certain types of environmental changes. These records are then extracted to construct temperature and humidity time series reflecting the environmental changes that led to the prediction failure. To compare and utilize these event data occurring at different times, the temperature and humidity gradients are calculated and time-aligned using a dynamic time warping algorithm. This time alignment eliminates differences in the speed or timing of environmental changes between different events, allowing environmental conditions with similar change patterns to be effectively correlated. The aligned environmental gradient series is used as input, and the corresponding uncompensated frequency drift is output. The prediction model parameters are optimized using a gradient descent algorithm. This process is equivalent to retraining or fine-tuning the prediction model using actual performance data from "difficult" examples (i.e., cases where inaccurate predictions resulted in frequency drift). By learning the relationship between the environmental change patterns and actual drift in these failure cases, the model improves its sensitivity and prediction accuracy to similar sudden or nonlinear environmental changes. As a result, the prediction model can more accurately predict potential future frequency drift, reduce the number of cases where the predicted drift exceeds the maximum adjustment range, and improve the effectiveness of overall frequency adjustment, thereby maintaining stable operation of the RF front-end control system.

[0081] In some specific embodiments, a controller in an RF front-end control system executes the above method. The controller includes a storage module for implementing an out-of-range drift database. A preset threshold for the number of out-of-range drift records is set, for example, to 20. When the number of records in the out-of-range drift database reaches 20, the controller initiates a model adjustment process. All records are read from the database. Each record includes a timestamp, temperature value, humidity value, and the corresponding frequency drift amount that exceeds the adjustment range. For each record, historical temperature and humidity data for a period of time prior to its timestamp is extracted to construct a temperature change sequence and a humidity change sequence. For example, temperature and humidity data for each second within 1 minute prior to the timestamp are extracted to form a sequence. First-order differences are calculated for each sequence to obtain a temperature change gradient sequence and a humidity change gradient sequence. For example, the temperature difference between adjacent time points is calculated as the temperature gradient. A dynamic time warping algorithm is used to pairwise align the temperature change gradient sequences corresponding to all records, and to pairwise align the humidity change gradient sequences, to obtain a set of aligned sequences. The dynamic time warping algorithm uses Euclidean distance to construct a cost matrix and employs a dynamic programming algorithm to find the optimal warping path. Based on the optimal path, the sequence is time-adjusted by repeating or skipping time points. The aligned temperature and humidity gradient sequences are used as input data sets, and the frequency drift in each record that exceeds the adjustment range is used as the output data set. If the prediction model is a neural network model, the Adam optimization algorithm is used as a gradient descent algorithm to adjust the neural network weights and bias parameters based on this data set. This optimizes the neural network model and enables more accurate prediction of frequency drift in similar rapid temperature and humidity change patterns.

[0082] In some embodiments, the steps of calculating a temperature change gradient sequence and a humidity change gradient sequence based on the temperature change sequence and the humidity change sequence, and performing time series alignment processing on the temperature change gradient sequence and the humidity change gradient sequence using a dynamic time warping algorithm to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence include:

[0083] For the temperature change sequence and humidity change sequence, the first-order difference calculation is used to obtain the temperature change gradient sequence and humidity change gradient sequence respectively;

[0084] Based on the pre-built cost matrix, a dynamic programming algorithm is used to find the optimal regularized path;

[0085] According to the optimal regularization path, the temperature change gradient sequence and the humidity change gradient sequence are time series aligned to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence; the alignment operation can be achieved by repeating or skipping certain time points to achieve the best match between the two sequences on the time axis.

[0086] For the temperature and humidity change sequences, first-order differences are used to generate temperature gradient sequences and humidity gradient sequences, respectively. First-order differences calculate the difference between values ​​at adjacent time points, thereby extracting information about the sequence's rate of change. The temperature gradient sequence reflects the speed and direction of temperature changes over time, while the humidity gradient sequence reflects the speed and direction of humidity changes over time. Extracting gradient information helps capture the dynamic nature of environmental changes. Furthermore, a dynamic programming algorithm is used to find the optimal warping path based on a pre-constructed cost matrix. The cost matrix quantifies the matching cost between any two time points in the temperature and humidity gradient sequences. The dynamic programming algorithm searches for a path within this cost matrix that connects the time points in the two sequences and minimizes the total matching cost. This optimal warping path allows for nonlinear stretching or compression of the sequences along the time axis to find the best correspondence, overcoming differences in sequence length, sampling rate, or rate of change. Based on the optimal warping path, the temperature and humidity gradient sequences are time-series aligned. Alignment is achieved by repeating or skipping certain time points. The optimal regularization path indicates which time points in the two original sequences should correspond. If the path indicates that a time point in the original sequence corresponds to multiple locations in the aligned sequence, stretching is achieved by repeating the data at that time point. If the path indicates that a time point in the original sequence has no corresponding location, compression is achieved by skipping that time point. This nonlinear alignment based on the optimal path ensures that meaningful correspondences can be established in the aligned sequence even if the original sequences are not synchronized in time or change at different rates.

[0087] Specifically, this technical solution aims to address the problem that time series corresponding to different out-of-limit drift records may differ in length, sampling interval, or rate of change, making simple alignment methods unable to accurately capture the true correspondence between the series. First, temperature and humidity change series are extracted from the out-of-limit drift records. Then, first-order differences are calculated for these series to obtain gradient series reflecting the temperature and humidity change rates. These gradient series can more directly reflect the dynamic process of environmental change. Next, to compare and align the gradient series from different out-of-limit events, a cost matrix is ​​constructed that quantifies the difference or mismatch between the gradient values ​​at any pair of time points in the two series. Using a dynamic programming algorithm, a path from the start to the end of the sequence with the lowest total cost is searched on this cost matrix, which is the optimal warped path. This path defines the optimal nonlinear mapping relationship between the two original gradient series on the time axis. Finally, based on the found optimal warped path, the original temperature and humidity gradient series are time-series aligned. Depending on the path indicated, the alignment process may require repeating or skipping data at certain time points in the original sequence, ensuring that the two sequences have the same length on the aligned timeline and that the data at corresponding positions have the highest similarity or lowest matching cost. This converts the original, potentially temporally divergent temperature and humidity change sequences into time-aligned gradient sequences. These aligned gradient sequences more accurately reflect the environmental variation patterns that lead to over-limit drift, providing high-quality input data for subsequent gradient descent algorithm adjustments of prediction model parameters. This improves the accuracy and robustness of the prediction model, ultimately enhancing the precision and effectiveness of FBAR filter resonant frequency adjustment.

[0088] In some specific embodiments, assume that temperature change sequences are extracted from two different out-of-limit drift records. After first-order difference calculation, two temperature change gradient sequences are obtained, for example, sequence A = [1.2, 0.8, 1.5] and sequence B = [0.9, 1.1, 0.7, 1.3]. These two sequences are of different lengths. To align them, a cost matrix is ​​constructed, where element C(i, j) represents the matching cost between the i-th element of sequence A and the j-th element of sequence B. For example, the absolute difference |A[i] - B[j]| can be used. Then, a dynamic programming algorithm is used to find an optimal regularized path from C(1,1) to C(3,4) on this cost matrix, with the minimum cumulative cost. For example, the optimal path found may indicate that the first point of sequence A corresponds to the first point of sequence B, the second point of sequence A corresponds to the second and third points of sequence B, and the third point of sequence A corresponds to the fourth point of sequence B. Time series alignment is performed based on this optimal path. The second point (0.8) of sequence A will be repeated once, resulting in the aligned sequence A' = [1.2, 0.8, 0.8, 1.5]. Sequence B remains as is or is adjusted according to the path, for example, the aligned sequence B' = [0.9, 1.1, 0.7, 1.3]. In this way, the two sequences are stretched or compressed on the time axis, allowing them to be meaningfully compared and used as input for the adjustment of the prediction model. The humidity gradient sequence is also processed in a similar manner.

[0089] In some embodiments, the step of performing time series alignment on the temperature change gradient sequence and the humidity change gradient sequence according to the optimal regularization path to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence includes:

[0090] According to the optimal regularization path, the time points that need to be aligned in the temperature gradient sequence and the humidity gradient sequence are determined, and a time point mapping table is constructed; the time point mapping table records the corresponding position of each time point in the temperature gradient sequence and the humidity gradient sequence in the aligned sequence;

[0091] According to the time point mapping table, the temperature change gradient sequence and the humidity change gradient sequence are adjusted in time series, specifically including: if there is a time point in the time point mapping table that corresponds to multiple time points, then linear interpolation is used for interpolation processing; if there is a time point in the time point mapping table that has no corresponding time point, then the time point is deleted from the sequence;

[0092] The adjusted temperature change gradient sequence and humidity change gradient sequence are spliced ​​to obtain aligned temperature change gradient sequence and humidity change gradient sequence.

[0093] After the optimal warping path is calculated according to the dynamic time warping algorithm, it is necessary to determine which time points in the original temperature change gradient sequence and the humidity change gradient sequence correspond to which positions in the aligned sequence based on the path. Specifically, each point on the optimal warping path is traversed, and the point consists of the index of the original temperature sequence and the index of the original humidity sequence. The length of the path determines the length of the aligned sequence. When constructing the time point mapping table, all positions where each index in the original sequence appears in the aligned sequence are recorded. For example, if the kth point on the optimal path corresponds to the index i of the original temperature sequence and the index j of the original humidity sequence, then in the time point mapping table of the temperature sequence, index i is recorded as the kth position of the corresponding aligned sequence; in the time point mapping table of the humidity sequence, index j is recorded as the kth position of the corresponding aligned sequence. After completing the construction of the mapping table, the original sequence is adjusted according to the mapping table. For a time point in the original sequence, if it corresponds to multiple aligned positions in the mapping table, it means that the time point is "stretched" on the time axis. At this time, the linear interpolation method is used to calculate the interpolation values ​​corresponding to multiple aligned positions based on the value of the time point and the values ​​of its adjacent time points. If a time point in the original sequence does not have a corresponding aligned position in the mapping table, it means that the time point is "compressed" or skipped on the time axis. At this time, the time point is deleted from the sequence. Through these adjustments, the original sequence is converted into an intermediate sequence of the same length that is consistent with the optimal regularization path. Finally, the adjusted temperature change gradient sequence and humidity change gradient sequence are combined at each time point to form the final aligned sequence. Each element of the aligned sequence contains the temperature gradient value and humidity gradient value of the corresponding time point.

[0094] Specifically, to address the problem that the dynamic time warping algorithm, after aligning time series of different lengths, directly generates a sequence structure that is unsuitable for prediction model input, this solution provides a method for converting the original gradient sequence into an aligned sequence with a unified structure. First, the optimal warping path calculated by the dynamic time warping algorithm reflects the optimal temporal correspondence between the two sequences. Based on this path, a time point mapping table is constructed, which explicitly records the corresponding position of each time point in the original temperature gradient sequence and the humidity gradient sequence in the aligned sequence. The length of the aligned sequence is equal to the length of the optimal warping path. Next, the original temperature gradient sequence and the humidity gradient sequence are adjusted based on the constructed time point mapping table. If the mapping table indicates that a time point in the original sequence corresponds to multiple positions in the aligned sequence, for example, a value in the original temperature gradient sequence needs to correspond to two consecutive positions in the aligned sequence, linear interpolation is used to calculate the interpolated values ​​at these two aligned positions based on the original value and its next value in the original sequence, thereby smoothly filling the aligned sequence. If the mapping table indicates that a time point in the original sequence has no corresponding aligned position, the original time point is skipped and deleted when constructing the adjusted sequence. This process converts the original sequence into an adjusted sequence that is precisely aligned with the optimal regularized path and has the same length. Finally, the adjusted temperature gradient sequence and humidity gradient sequence are concatenated to form an aligned sequence containing paired temperature and humidity gradient values. This aligned sequence, used as input to the subsequent prediction model, more accurately reflects the combined dynamic characteristics of temperature and humidity variations, thereby improving the accuracy of predicting the resonant frequency drift of the FBAR filter.

[0095] In some specific embodiments, it is assumed that the original temperature change gradient sequence is T=[1,2,3], and the original humidity change gradient sequence is H=[10,20]. The optimal warping path obtained by dynamic time warping calculation is P=[(0,0),(1,0),(1,1),(2,1)]. The path length is 4, so the length of the aligned sequence is 4. A time point mapping table is constructed based on the path: the original T[0] corresponds to alignment position 0; the original T[1] corresponds to alignment positions 1 and 2; the original T[2] corresponds to alignment position 3. The original H[0] corresponds to alignment positions 0 and 1; the original H[1] corresponds to alignment positions 2 and 3. Then the sequence is adjusted. For the temperature sequence T: T[0]=1, corresponding to alignment position 0, after adjustment T'[0]=1. T[1]=2, corresponding to alignment positions 1 and 2. Linear interpolation is used, and the interpolation range is between T[1]=2 and T[2]=3, and is distributed to two positions. After adjustment, T'[1]=T[1]+(T[2]-T[1])*((1-1) / (3-1))=2+1*0=2. After adjustment, T'[2]=T[1]+(T[2]-T[1])*((2-1) / (3-1))=2+1*0.5=2.5. T[2]=3, corresponding to alignment position 3, after adjustment, T'[3]=3. The adjusted temperature change gradient sequence is T'=[1,2,2.5,3]. For the humidity sequence H: H[0]=10, corresponding to alignment positions 0 and 1. Linear interpolation is used, and the interpolation range is between H[0]=10 and H[1]=20, distributed to two positions. After adjustment, H'[0]=H[0]+(H[1]-H[0])*((0-0) / (2-0))=10+10*0=10. After adjustment, H'[1]=H[0]+(H[1]-H[0])*((1-0) / (2-0))=10+10*0.5=15. H[1]=20, corresponding to alignment positions 2 and 3. Since H[1] is the last point and corresponds to multiple positions, simple repetition or other interpolation strategies can be used. Assuming simple repetition, after adjustment, H'[2]=20, H'[3]=20. The adjusted humidity change gradient sequence is H'=[10,15,20,20]. Finally, the adjusted T' and H' are spliced ​​to obtain the aligned sequence, for example, represented as a series of vectors: [(1,10),(2,15),(2.5,20),(3,20)]. This aligned sequence serves as input data for the prediction model. This approach allows for the generation of uniformly structured input data that reflects synchronized changes, even when the original temperature and humidity gradient sequences are of varying lengths. This improves the training efficiency and prediction accuracy of the prediction model.

[0096] In certain embodiments, the step of using the aligned temperature change gradient sequence and humidity change gradient sequence as input and the frequency drift amount exceeding the adjustment range in each out-of-limit drift record as output, and adjusting the model parameters of the prediction model using a gradient descent algorithm includes:

[0097] Determine the type of prediction model, specifically: if the prediction model is a neural network model, use the Adam optimization algorithm as the gradient descent algorithm; if the prediction model is a support vector machine model, use the SMO algorithm as the gradient descent algorithm;

[0098] According to the ambient temperature data and ambient humidity data in the over-limit drift record, the learning rate adjustment range is determined by calculating the temperature change range and the humidity change range;

[0099] The aligned temperature change gradient sequence and humidity change gradient sequence are taken as input, and the frequency drift amount exceeding the adjustment range in each out-of-limit drift record is taken as output. The Adam optimization algorithm or the SMO algorithm is used to adaptively adjust the learning rate within the learning rate adjustment range, and the model parameters of the prediction model are adjusted based on the adjusted learning rate.

[0100] In this embodiment, the efficiency of parameter updating can be improved by selecting a suitable optimization algorithm based on the structural characteristics of the model. Furthermore, the learning rate adjustment range is determined by calculating the temperature variation range and the humidity variation range based on the ambient temperature data and the ambient humidity data in the over-limit drift record. This range provides a boundary for the subsequent adaptive adjustment of the learning rate. Specifically, the aligned temperature variation gradient sequence and the humidity variation gradient sequence are used as input, and the frequency drift amount exceeding the adjustment range in each over-limit drift record is used as output. The Adam optimization algorithm or the SMO algorithm is used to adaptively adjust the learning rate within the learning rate adjustment range, and the model parameters of the prediction model are adjusted based on the adjusted learning rate. The adaptive adjustment of the learning rate enables the optimization process to dynamically adjust the step size according to the current training state and data characteristics, thereby more effectively finding the optimal model parameters.

[0101] Specifically, this technical solution addresses the problem of slow convergence or local optima caused by a fixed learning rate failing to adapt to complex temperature and humidity variations during prediction model parameter adjustment. First, an optimization algorithm is selected based on the type of prediction model being optimized, such as the Adam algorithm for neural network models and the SMO algorithm for support vector machine models. This ensures the compatibility of the optimization algorithm with the model structure. Next, the collected out-of-range drift records are analyzed to calculate the range of variation in the ambient temperature and humidity data. For example, the maximum temperature change or the maximum humidity change during the recording period can be calculated. Based on this calculated range, an appropriate learning rate adjustment range is determined. This range sets a limit for subsequent adaptive learning rate adjustments, preventing the learning rate from being too large or too small. The time-aligned temperature and humidity gradient series are then used as the model input data, and the actual frequency drift outside the adjustment range in each out-of-range drift record is used as the model output target. The model parameters are updated using the previously determined optimization algorithm (Adam or SMO) within the set learning rate adjustment range. During this process, the optimization algorithm adaptively adjusts the learning rate based on current gradient information and historical update information. For example, the Adam algorithm maintains an independent adaptive learning rate for each model parameter. This adaptive adjustment allows the parameter update step size to be dynamically adjusted based on the local characteristics of the loss function, thereby more effectively traversing the complex loss function surface and avoiding stagnation in flat areas or oscillation in steep areas. Ultimately, through iterative optimization, the model parameters are adjusted to better fit the relationship between the input (temperature and humidity gradient sequence) and the output (excessive frequency drift). This improves the accuracy of the prediction model, enabling more precise prediction of the future resonant frequency drift of the FBAR filter, thereby improving the precision of resonant frequency adjustment.

[0102] In some specific embodiments, the prediction model is assumed to be a feedforward neural network used to predict out-of-limit frequency drift. After a certain number (e.g., 100) of out-of-limit drift records have been collected, the model parameter adjustment process is initiated. First, the prediction model is determined to be a neural network, and therefore the Adam optimization algorithm is selected. Next, ambient temperature and humidity data are extracted from these 100 out-of-limit drift records. The maximum absolute value of the temperature change (e.g., the difference between the highest and lowest temperatures) and the maximum absolute value of the humidity change across all records are calculated. Based on these maximum change values, the initial learning rate and learning rate decay strategy of the Adam optimizer are set, or a learning rate upper and lower limit range is determined, for example, setting the learning rate adjustment range between 0.0005 and 0.05. Then, the aligned temperature change gradient sequence and humidity change gradient sequence corresponding to each out-of-limit drift record are used as input features of the neural network, and the frequency drift amount in the record that exceeds the adjustment range is used as the training target. The neural network is trained using the Adam optimization algorithm. During the training process, the Adam algorithm adaptively adjusts the learning rate of each parameter based on its gradient information to ensure efficient parameter updates. For example, for parameters with large gradient changes, the learning rate might be relatively large; for parameters with small gradient changes, the learning rate might be relatively small. This adaptive process occurs within a set learning rate adjustment range. Through multiple training cycles, the neural network's weights and bias parameters are optimized, enabling the model to more accurately predict out-of-limit frequency drift that may occur under specific temperature and humidity variation patterns.

[0103] The present invention provides a radio frequency front-end control system, comprising a controller, which is used to execute the steps in the frequency adjustment method based on the SIP module in the above embodiment.

[0104] Please refer to Figure 2 , Figure 2 In some embodiments of the present invention, an adjustment device based on a SIP module is applied to a radio frequency front-end control system. The adjustment device based on the SIP module is integrated into a back-end control device in the form of a computer program, and includes:

[0105] The acquisition module 100 is used to acquire ambient temperature data and ambient humidity data, and calculate the temperature change rate and humidity change rate based on the ambient temperature data and ambient humidity data;

[0106] The prediction module 200 is used to input the temperature change rate and the humidity change rate into a preset prediction model to predict the future resonant frequency drift of the FBAR filter;

[0107] The calculation module 300 is used to calculate the resonant frequency adjustment amount according to the predicted resonant frequency drift amount;

[0108] A generating module 400 is configured to generate a control signal based on a resonant frequency adjustment amount according to a response characteristic and an adjustment step size of an adjustable element on the FBAR filter;

[0109] The control module 500 is used to control the adjustable elements on the FBAR filter to change the filter parameters using a control signal, so as to adjust the resonant frequency of the FBAR filter.

[0110] In some embodiments, when calculating the resonant frequency adjustment amount according to the predicted resonant frequency drift amount, the calculation module 300 executes:

[0111] Determining a maximum resonant frequency adjustment amount based on an initial resonant frequency of the FBAR filter and a preset frequency adjustment range;

[0112] Determining whether the predicted resonant frequency drift exceeds an adjustment range limited by a maximum resonant frequency adjustment amount;

[0113] If the predicted resonant frequency drift does not exceed the adjustment range, the predicted resonant frequency drift is used as the resonant frequency adjustment amount;

[0114] If the predicted resonant frequency drift exceeds the adjustment range, the resonant frequency adjustment amount is set to the maximum resonant frequency adjustment amount, and the frequency drift amount exceeding the adjustment range, the current ambient temperature data, the current ambient humidity data, and the timestamp information are encapsulated as an out-of-limit drift record; the out-of-limit drift record is used to adjust the model parameters of the prediction model.

[0115] In some embodiments, the adjustment device based on the SIP module further includes an optimization module 600, which is configured to perform the following steps:

[0116] Adding the over-limit drift record to the over-limit drift database, and determining whether the number of over-limit drift records in the over-limit drift database exceeds a preset over-limit drift record number threshold;

[0117] If the number of out-of-limit drift records in the out-of-limit drift database exceeds a preset threshold number of out-of-limit drift records, all out-of-limit drift records are extracted from the out-of-limit drift database, and a temperature change sequence and a humidity change sequence are constructed based on the ambient temperature data, ambient humidity data, and timestamp information in each out-of-limit drift record;

[0118] According to the temperature change sequence and the humidity change sequence, the temperature change gradient sequence and the humidity change gradient sequence are calculated, and the temperature change gradient sequence and the humidity change gradient sequence are aligned using the dynamic time warping algorithm to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence;

[0119] The aligned temperature change gradient sequence and humidity change gradient sequence are taken as input, and the frequency drift amount exceeding the adjustment range in each over-limit drift record is taken as output. The gradient descent algorithm is used to adjust the model parameters of the prediction model.

[0120] In certain embodiments, when the optimization module 600 is used to calculate a temperature change gradient sequence and a humidity change gradient sequence based on a temperature change sequence and a humidity change sequence, and to perform time series alignment processing on the temperature change gradient sequence and the humidity change gradient sequence using a dynamic time warping algorithm to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence, the following is executed:

[0121] For the temperature change sequence and humidity change sequence, the first-order difference calculation is used to obtain the temperature change gradient sequence and humidity change gradient sequence respectively;

[0122] Based on the pre-built cost matrix, a dynamic programming algorithm is used to find the optimal regularized path;

[0123] According to the optimal regularization path, the temperature change gradient sequence and the humidity change gradient sequence are time series aligned to obtain the aligned temperature change gradient sequence and the aligned humidity change gradient sequence.

[0124] In some embodiments, when the optimization module 600 is used to perform time series alignment on the temperature change gradient sequence and the humidity change gradient sequence according to the optimal regularization path to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence, the following steps are executed:

[0125] According to the optimal regularization path, the time points that need to be aligned in the temperature gradient sequence and the humidity gradient sequence are determined, and a time point mapping table is constructed; the time point mapping table records the corresponding position of each time point in the temperature gradient sequence and the humidity gradient sequence in the aligned sequence;

[0126] According to the time point mapping table, the temperature change gradient sequence and the humidity change gradient sequence are adjusted in time series, specifically including: if there is a time point in the time point mapping table that corresponds to multiple time points, then linear interpolation is used for interpolation processing; if there is a time point in the time point mapping table that has no corresponding time point, then the time point is deleted from the sequence;

[0127] The adjusted temperature change gradient sequence and humidity change gradient sequence are spliced ​​to obtain aligned temperature change gradient sequence and humidity change gradient sequence.

[0128] In certain embodiments, the optimization module 600 is configured to use the aligned temperature change gradient sequence and humidity change gradient sequence as input, and output the frequency drift amount exceeding the adjustment range in each out-of-limit drift record, and to adjust the model parameters of the prediction model using a gradient descent algorithm. The following steps are executed:

[0129] Determine the type of prediction model, specifically: if the prediction model is a neural network model, use the Adam optimization algorithm as the gradient descent algorithm; if the prediction model is a support vector machine model, use the SMO algorithm as the gradient descent algorithm;

[0130] According to the ambient temperature data and ambient humidity data in the over-limit drift record, the learning rate adjustment range is determined by calculating the temperature change range and the humidity change range;

[0131] The aligned temperature change gradient sequence and humidity change gradient sequence are taken as input, and the frequency drift amount exceeding the adjustment range in each out-of-limit drift record is taken as output. The Adam optimization algorithm or the SMO algorithm is used to adaptively adjust the learning rate within the learning rate adjustment range, and the model parameters of the prediction model are adjusted based on the adjusted learning rate.

[0132] Please refer to Figure 3 , Figure 3 This is a structural diagram of an electronic device provided by an embodiment of the present invention. The present invention provides an electronic device 13, including: a processor 1301 and a memory 1302. The processor 1301 and the memory 1302 are interconnected and communicate with each other via a communication bus 1303 and / or other forms of connection mechanisms (not shown). The memory 1302 stores computer-readable instructions executable by the processor 1301. When the electronic device is running, the processor 1301 executes the computer-readable instructions to execute the frequency adjustment method based on the SIP module in any optional implementation of the above embodiment to achieve the following functions: collecting ambient temperature The method comprises the following steps: collecting the ambient temperature data and the ambient humidity data, and calculating the temperature change rate and the humidity change rate based on the ambient temperature data and the ambient humidity data; inputting the temperature change rate and the humidity change rate into a preset prediction model to predict the future resonant frequency drift of the FBAR filter; establishing the prediction model based on the historical temperature and humidity change data; calculating the resonant frequency adjustment amount based on the predicted resonant frequency drift; generating a control signal based on the resonant frequency adjustment amount according to the response characteristics and adjustment step size of the adjustable element on the FBAR filter; and utilizing the control signal to control the adjustable element on the FBAR filter to change the filter parameters so as to adjust the resonant frequency of the FBAR filter.

[0133] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the frequency adjustment method based on the SIP module in any optional implementation of the above-mentioned embodiment is executed to achieve the following functions: collecting ambient temperature data and ambient humidity data, and calculating the temperature change rate and the humidity change rate based on the ambient temperature data and the ambient humidity data; inputting the temperature change rate and the humidity change rate into a preset prediction model to predict the future resonant frequency drift of the FBAR filter; establishing the prediction model based on historical temperature and humidity change data; calculating the resonant frequency adjustment amount based on the predicted resonant frequency drift; generating a control signal based on the resonant frequency adjustment amount according to the response characteristics and adjustment step size of the adjustable element on the FBAR filter; and using the control signal to control the adjustable element on the FBAR filter to change the filter parameters to adjust the resonant frequency of the FBAR filter.

[0134] Among them, the computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0135] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some communication interface, the indirect coupling or communication connection of the device or unit may be electrical, mechanical or other forms.

[0136] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0137] Furthermore, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0138] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any actual relationship or order between these entities or operations.

[0139] The foregoing description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A frequency adjustment method based on a sip module, applied to a radio frequency front-end control system, characterized in that: Including steps: Collecting ambient temperature data and ambient humidity data, and calculating the temperature change rate and humidity change rate based on the ambient temperature data and ambient humidity data; Inputting the temperature change rate and the humidity change rate into a preset prediction model to predict the future resonant frequency drift of the FBAR filter; the prediction model is established based on historical temperature and humidity change data; Calculating a resonant frequency adjustment amount based on the predicted resonant frequency drift amount; generating a control signal based on the resonant frequency adjustment amount according to the response characteristics and adjustment step size of the adjustable element of the FBAR filter; Using the control signal, controlling the adjustable element on the FBAR filter to change the filter parameters to adjust the resonant frequency of the FBAR filter; The steps of calculating the resonant frequency adjustment amount according to the predicted resonant frequency drift amount include: Determining a maximum resonant frequency adjustment amount based on an initial resonant frequency of the FBAR filter and a preset frequency adjustment range; Determining whether the predicted resonant frequency drift exceeds an adjustment range defined by the maximum resonant frequency adjustment amount; If the predicted resonant frequency drift does not exceed the adjustment range, the predicted resonant frequency drift is used as the resonant frequency adjustment amount; If the predicted resonant frequency drift exceeds the adjustment range, the resonant frequency adjustment amount is set to the maximum resonant frequency adjustment amount, and the frequency drift amount exceeding the adjustment range, the current ambient temperature data, the current ambient humidity data, and the timestamp information are encapsulated into an out-of-limit drift record; the out-of-limit drift record is used to adjust the model parameters of the prediction model; The frequency adjustment method based on the SIP module also includes the following steps: Adding the over-limit drift record to an over-limit drift database, and determining whether the number of over-limit drift records in the over-limit drift database exceeds a preset over-limit drift record number threshold; If the number of out-of-limit drift records in the out-of-limit drift database exceeds a preset threshold number of out-of-limit drift records, all out-of-limit drift records are extracted from the out-of-limit drift database, and a temperature change sequence and a humidity change sequence are constructed based on the ambient temperature data, ambient humidity data, and timestamp information in each out-of-limit drift record; Calculating a temperature change gradient sequence and a humidity change gradient sequence according to the temperature change sequence and the humidity change sequence, and performing time series alignment processing on the temperature change gradient sequence and the humidity change gradient sequence using a dynamic time warping algorithm to obtain aligned temperature change gradient sequence and humidity change gradient sequence; The aligned temperature change gradient sequence and humidity change gradient sequence are used as input, and the frequency drift amount exceeding the adjustment range in each over-limit drift record is used as output, and the model parameters of the prediction model are adjusted using a gradient descent algorithm.

2. The frequency adjustment method based on the SIP module according to claim 1, characterized in that: The steps of calculating a temperature change gradient sequence and a humidity change gradient sequence according to the temperature change sequence and the humidity change sequence, and performing time series alignment processing on the temperature change gradient sequence and the humidity change gradient sequence using a dynamic time warping algorithm to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence include: For the temperature change sequence and the humidity change sequence, first-order difference calculation is used to obtain a temperature change gradient sequence and a humidity change gradient sequence respectively; Based on the pre-built cost matrix, a dynamic programming algorithm is used to find the optimal regularized path; According to the optimal regularization path, the temperature change gradient sequence and the humidity change gradient sequence are time-series aligned to obtain aligned temperature change gradient sequence and humidity change gradient sequence.

3. The frequency adjustment method based on the SIP module according to claim 2, characterized in that: According to the optimal regularization path, the temperature change gradient sequence and the humidity change gradient sequence are time-series aligned to obtain the aligned temperature change gradient sequence and humidity change gradient sequence, comprising: Determine the time points that need to be aligned in the temperature change gradient sequence and the humidity change gradient sequence according to the optimal regularization path, and construct a time point mapping table; the time point mapping table records the corresponding position of each time point in the temperature change gradient sequence and the humidity change gradient sequence in the aligned sequence; According to the time point mapping table, the temperature change gradient sequence and the humidity change gradient sequence are adjusted in time series, specifically including: if there is a time point in the time point mapping table that corresponds to multiple time points, interpolation processing is performed using a linear interpolation method; if there is a time point in the time point mapping table that has no corresponding time point, deleting the time point from the sequence; The adjusted temperature change gradient sequence and humidity change gradient sequence are spliced ​​to obtain aligned temperature change gradient sequence and humidity change gradient sequence.

4. The frequency adjustment method based on the SIP module according to claim 1, characterized in that: The steps of using the aligned temperature change gradient sequence and humidity change gradient sequence as input, and outputting the frequency drift amount exceeding the adjustment range in each over-limit drift record, and adjusting the model parameters of the prediction model using a gradient descent algorithm include: Determine the type of prediction model, specifically: if the prediction model is a neural network model, use the Adam optimization algorithm as the gradient descent algorithm; if the prediction model is a support vector machine model, use the SMO algorithm as the gradient descent algorithm; According to the ambient temperature data and ambient humidity data in the over-limit drift record, the learning rate adjustment range is determined by calculating the temperature change range and the humidity change range; The aligned temperature change gradient sequence and humidity change gradient sequence are taken as input, and the frequency drift amount exceeding the adjustment range in each out-of-limit drift record is taken as output. The Adam optimization algorithm or the SMO algorithm is used to adaptively adjust the learning rate within the learning rate adjustment range, and the model parameters of the prediction model are adjusted based on the adjusted learning rate.

5. A radio frequency front-end control system, characterized in that: It includes a controller, which is used to execute the steps in the frequency adjustment method based on the SIP module as described in any one of claims 1 to 4.

6. An adjustment device based on a sip module, applied to a radio frequency front-end control system, characterized in that: include: The acquisition module is used to collect ambient temperature data and ambient humidity data, and calculate the temperature change rate and humidity change rate based on the ambient temperature data and ambient humidity data; A prediction module is used to input the temperature change rate and the humidity change rate into a preset prediction model to predict the future resonant frequency drift of the FBAR filter; A calculation module, configured to calculate a resonant frequency adjustment amount based on the predicted resonant frequency drift amount; A generating module, configured to generate a control signal based on the resonant frequency adjustment amount according to a response characteristic and an adjustment step size of an adjustable element on the FBAR filter; a control module, configured to use the control signal to control an adjustable element on the FBAR filter to change filter parameters, thereby adjusting the resonant frequency of the FBAR filter; The calculation module is used to calculate the resonant frequency adjustment amount according to the predicted resonant frequency drift amount when executing: Determining a maximum resonant frequency adjustment amount based on an initial resonant frequency of the FBAR filter and a preset frequency adjustment range; Determining whether the predicted resonant frequency drift exceeds an adjustment range limited by a maximum resonant frequency adjustment amount; If the predicted resonant frequency drift does not exceed the adjustment range, the predicted resonant frequency drift is used as the resonant frequency adjustment amount; If the predicted resonant frequency drift exceeds the adjustment range, the resonant frequency adjustment amount is set to the maximum resonant frequency adjustment amount, and the frequency drift exceeding the adjustment range, the current ambient temperature data, the current ambient humidity data, and the timestamp information are encapsulated into an out-of-limit drift record; the out-of-limit drift record is used to adjust the model parameters of the prediction model; The adjustment device based on the SIP module also includes an optimization module, which is used to perform the following steps: Adding the over-limit drift record to the over-limit drift database, and determining whether the number of over-limit drift records in the over-limit drift database exceeds a preset over-limit drift record number threshold; If the number of out-of-limit drift records in the out-of-limit drift database exceeds a preset threshold number of out-of-limit drift records, all out-of-limit drift records are extracted from the out-of-limit drift database, and a temperature change sequence and a humidity change sequence are constructed based on the ambient temperature data, ambient humidity data, and timestamp information in each out-of-limit drift record; According to the temperature change sequence and the humidity change sequence, the temperature change gradient sequence and the humidity change gradient sequence are calculated, and the temperature change gradient sequence and the humidity change gradient sequence are aligned using the dynamic time warping algorithm to obtain the aligned temperature change gradient sequence and the humidity change gradient sequence; The aligned temperature change gradient sequence and humidity change gradient sequence are taken as input, and the frequency drift amount exceeding the adjustment range in each over-limit drift record is taken as output. The gradient descent algorithm is used to adjust the model parameters of the prediction model.

7. An electronic device, characterized in that: The system comprises a processor and a memory, wherein the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the steps of the frequency adjustment method based on the SIP module according to any one of claims 1 to 4 are executed.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the frequency adjustment method based on the SIP module are executed.

Images