A frequency adaptive stabilization system for a signal shielding device

Abstract

The application discloses a signal shielding equipment frequency adaptive stabilization system, and relates to the technical field of radio frequency signal processing and electromagnetic spectrum management. The system generates a shielding signal through a radio frequency signal generation unit, and a frequency error detection unit adopts an independent local oscillator and a digital phase discrimination architecture to quantize the frequency offset of a radio frequency output signal in real time. Meanwhile, a multi-source sensing unit collects temperature, voltage and running time information, and an intelligent decision and control unit fuses the above multi-dimensional data based on an extended Kalman filter to separate the frequency offset components caused by temperature drift, aging and power disturbance, and calculates the optimal compensation control amount through a model predictive controller. Finally, a radio frequency front-end execution unit executes frequency compensation in a combination of coarse adjustment and fine adjustment. The application constructs a complete closed-loop adaptive architecture, and high-precision frequency locking can be realized without high-cost constant temperature crystal oscillators.

Description

Technical Field

[0001] This invention relates to the fields of radio frequency signal processing and electromagnetic spectrum management, and in particular to a frequency adaptive stabilization system for signal shielding devices. Background Technology

[0002] Signal jamming technology is widely used in scenarios such as secure meetings, exam cheating prevention, prison management, and communication countermeasures. With the widespread adoption of digital signaling technologies such as 5G New Radio (5G NR) and narrowband IoT, communication systems face extremely high requirements for carrier frequency locking accuracy. Taking 5G NR systems as an example, their subcarrier spacing can be as narrow as 15kHz, and the Automatic Frequency Control (AFC) mechanism can complete synchronization and demodulation within a frequency offset range of ±0.1 ppm to ±2 ppm. If the frequency offset of the jamming signal exceeds this range, the communication equipment can avoid interference, leading to jamming failure.

[0003] Signal shielding devices typically employ a frequency synthesis scheme using a temperature-compensated crystal oscillator (TCC) in conjunction with a phase-locked loop (PLL) and a voltage-controlled oscillator (VCO). However, under conditions of ambient temperature variations and internal power amplifier self-heating, the output frequency can experience significant drift. Several existing technologies address this frequency drift problem.

[0004] Passive cooling combined with factory static calibration involves reducing temperature rise through heat dissipation via heat sinks, fans, and a metal casing, and storing a temperature-frequency correction table before shipment for runtime adjustments. The drawback of this method is that it cannot handle nonlinear dynamic temperature drift and crystal oscillator aging. Furthermore, factory calibration data may not remain valid during long-term operation or when the device is in a miniaturized, high-power emission state.

[0005] Wideband coverage strategy involves significantly expanding the transmit bandwidth to mask frequency offset. The drawback of this approach is that, in a digital signaling system, frequency offset can cause the shielded signal energy to fail to accurately fall into the target resource block, making it easier for the receiver's AFC mechanism to identify and filter it out. Furthermore, this approach sacrifices adjacent channel selectivity and results in power waste.

[0006] The open-loop temperature compensation scheme involves deploying temperature sensors near key components and adjusting the VCO tuning voltage or PLL division ratio after obtaining the frequency deviation estimate through table lookup or formula fitting. The drawbacks of this approach are that temperature sensing and frequency execution are independent, making it impossible to directly quantify the actual frequency error at the RF output, distinguish whether the frequency deviation originates from temperature drift, component aging, or power supply fluctuations, and the compensation effect cannot be verified.

[0007] In summary, existing technologies generally suffer from the following shortcomings: a lack of real-time online direct detection capability for frequency offset; failure to construct an integrated closed-loop control architecture for sensing, detection, and compensation; and difficulty in meeting the stringent requirements of signaling-level shielding for precise frequency locking and high narrowband selectivity. Furthermore, while temperature-controlled crystal oscillators can provide high frequency stability, their high cost makes them unsuitable for use in cost-sensitive shielding equipment. Summary of the Invention

[0008] (a) Purpose of the invention

[0009] To address at least one of the problems existing in the prior art, the present invention provides an adaptive frequency stabilization system that can detect the actual frequency offset of the RF output terminal in real time and organically combine frequency error detection, multi-source environmental perception and intelligent closed-loop compensation. The system aims to achieve long-term stability and precise locking of the shielded signal frequency without relying on high-cost temperature-controlled crystal oscillators.

[0010] (II) Technical Solution

[0011] To achieve the above-mentioned objectives, the present invention provides a frequency adaptive stabilization system for a signal shielding device, comprising a radio frequency signal generation unit, a frequency error detection unit, a multi-source sensing unit, an intelligent decision and control unit, and a radio frequency front-end execution unit, wherein the above units form a complete negative feedback closed loop.

[0012] Radio frequency signal generation unit:

[0013] The components used to generate the shielding signal at the desired frequency include a crystal reference source, a phase-locked loop, and a voltage-controlled oscillator. The crystal reference source can be a standard temperature-compensated crystal oscillator or a standard quartz crystal oscillator.

[0014] Frequency error detection unit:

[0015] A portion of the RF transmission signal is coupled from the output of the RF front-end execution unit via a directional coupler, and the frequency offset Δf of the RF output signal is detected and quantified in real time using a zero-difference frequency locking and digital phase detection architecture.

[0016] Specifically, the coupled signal is mixed with an independent local oscillator signal. This local oscillator uses an independent clock source from the RF signal generation unit and is physically isolated from it. The mixed signal is then filtered by a bandpass filter to extract the difference frequency signal. The difference frequency signal is shaped into a square wave by a comparator and fed into a counter within a Field-Programmable Gate Array (FPGA). A multi-cycle synchronous measurement method is used to calculate the actual frequency of the difference frequency signal, thereby deriving the real-time frequency deviation Δf of the RF output signal. This frequency error detection unit has a detection resolution better than 0.5 Hz and a refresh rate of 1 kHz.

[0017] Multi-source sensing unit:

[0018] Used to collect multidimensional environmental and state information related to frequency stability, including:

[0019] (1) Temperature information collected by temperature sensors deployed near the voltage-controlled oscillator, phase-locked loop chip, and crystal oscillator;

[0020] (2) Power supply voltage information monitored at the power supply input terminal of the radio frequency signal generation unit;

[0021] (3) System runtime information maintained within the intelligent decision and control unit.

[0022] Intelligent decision and control unit:

[0023] Receive the real-time frequency offset Δf from the frequency error detection unit and the multi-dimensional state information from the multi-source sensing unit, and perform the following processing:

[0024] First, an extended Kalman filter is used, with Δf as the observed quantity and temperature change, runtime, and voltage change as inputs. Based on the physical model describing the source of frequency shift, Δf = αΔT + βt + γΔV + w, the temperature drift component Δf is estimated and separated online. T , Aging component Δf A and power supply disturbance component Δf P , where α is the temperature drift coefficient, β is the aging drift rate, γ is the voltage sensitivity, and w is the process noise.

[0025] Second, based on the estimation results of the extended Kalman filter, a model predictive controller is adopted. The system model Δf(k + 1) = Δf(k) + u(k) + w(k) is used to predict the frequency offset trend over multiple control cycles, where k is the discrete time index of the current moment. The optimal compensation control quantity u is calculated by solving a constrained optimization problem. The cost function of this optimization problem is J = Σ(Q err ·Δf(k + i) 2 + R u ·u(k + i) 2 ), where i is the step index in the prediction time domain, Q err R is the penalty weight for frequency offset error. u Penalty weights are applied to changes in the control quantity.

[0026] Third, the recursive least squares method is used to update the device-specific temperature-frequency deviation model parameters and aging-frequency deviation model parameters online. The forgetting factor λ of the recursive least squares method is 0.99.

[0027] RF front-end execution unit:

[0028] The system receives a digital compensation control quantity *u* from the intelligent decision and control unit and performs frequency compensation using a hybrid execution strategy combining coarse and fine tuning. The fine tuning path outputs an analog voltage to the tuning voltage input of the voltage-controlled oscillator (VCO) via a digital-to-analog converter (DAC) to achieve fine frequency adjustment. The coarse tuning path achieves wide-range frequency coarse tuning by writing a division ratio correction value to the phase-locked loop (PLL) chip. When the detected frequency deviation Δf exceeds the linear tuning range of the fine tuning path, the coarse tuning path is used first to pull the frequency deviation back into the linear operating region of the fine tuning path, after which the fine tuning path performs fine tuning.

[0029] The workflow of the above system is as follows: the radio frequency signal generation unit generates a shielded signal and transmits it through the radio frequency front-end execution unit; the frequency error detection unit detects the current frequency offset Δf in real time and sends it to the intelligent decision and control unit; the multi-source sensing unit synchronously collects temperature, voltage and runtime information and sends it to the intelligent decision and control unit; the intelligent decision and control unit calculates the optimal compensation control quantity u through an extended Kalman filter and a model predictive controller and sends it to the radio frequency front-end execution unit; the radio frequency front-end execution unit performs frequency compensation by combining coarse and fine tuning; after frequency adjustment, the system immediately enters the next round of closed-loop iteration of detection and compensation.

[0030] The present invention also provides a control method for a frequency adaptive stabilization system of a signal shielding device, including the various steps performed by the above system.

[0031] (III) Beneficial Effects

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] (1) Real-time online direct detection of frequency offset. Through an independent local oscillator and digital phase detection architecture, the actual frequency error at the RF output can be directly quantified without relying on indirect inference from a temperature sensor. The detection resolution is better than 0.5 Hz, providing accurate and timely feedback for closed-loop compensation.

[0034] (2) Construct a complete closed-loop control architecture. Integrate frequency error detection, multi-source sensing, intelligent decision-making and RF front-end execution to form a negative feedback closed loop of sensing-detection-compensation, enabling the system to dynamically and adaptively perform real-time fine-tuning under operating conditions such as temperature changes, device aging or power fluctuations.

[0035] (3) Achieve intelligent identification and targeted compensation of frequency offset sources. By fusing multi-source data through an extended Kalman filter, the frequency offset components caused by temperature drift, aging and power supply disturbance can be separated online. The model predictive controller is used to make forward-looking optimal compensation decisions, and the compensation accuracy and response speed are better than the existing open-loop schemes.

[0036] (4) Reduce hardware costs. High-precision frequency locking can be achieved by using ordinary temperature-compensated crystal oscillators or ordinary quartz crystal oscillators without relying on high-cost temperature-compensated crystal oscillators.

[0037] (5) Improve spectrum selectivity and environmental adaptability. The system adopts narrowband transmission, which significantly reduces adjacent channel interference and has good frequency stability over a wide temperature range. It also has the ability to quickly suppress power fluctuations. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the overall architecture of the frequency adaptive stabilization system of the signal shielding device of the present invention.

[0039] Figure 2 This is a flowchart of the closed-loop control of the frequency adaptive stabilization system of the signal shielding device of the present invention. Detailed Implementation

[0040] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are for illustrative purposes only and should not be considered as limiting the scope of the invention. Those skilled in the art can make various modifications and equivalent substitutions to the technical solution of the present invention without departing from the principles of the invention, and all such modifications and substitutions fall within the protection scope of the present invention.

[0041] Example 1: System Hardware Deployment

[0042] This embodiment provides a specific hardware implementation of a frequency adaptive stabilization system for a signal shielding device. The overall system architecture is as follows: Figure 1 As shown. This embodiment takes a signaling-level jammer in the 5G NR n78 band (3.3-3.8 GHz) as an example, with the target signal being a 3.5 GHz carrier.

[0043] 1. Deployment of Radio Frequency Signal Generation Unit

[0044] The radio frequency signal generation unit, which is used to generate a shielded signal, includes a crystal reference source, a phase-locked loop and a voltage-controlled oscillator in claim 1.

[0045] A wideband fractional-N phase-locked loop (PLL) chip with an integrated VCO, the LMX2594, was selected, with a 50 MHz ordinary quartz crystal oscillator SG-210STF as the reference clock source. The PLL output frequency was set to 3.5 GHz. A third-order passive filter was preferably used for the loop filter, and the loop bandwidth was preferably set to 100 kHz.

[0046] 2. Deployment of Frequency Error Detection Unit

[0047] The "frequency error detection unit" in claim 1 is used to detect and quantify the frequency offset Δf of the RF output signal in real time using an independent local oscillator and a digital phase detector architecture, wherein the local oscillator and the RF signal generation unit use independent clock sources and are physically isolated from each other. This unit directly detects the actual frequency offset through an independent local oscillator, without relying on indirect inference based on temperature, thus providing accurate feedback for closed-loop compensation.

[0048] A miniature directional coupler ZFDC-20-5G is installed between the output of the RF front-end execution unit and the antenna. The preferred coupling is -20 dB, the isolation is greater than 45 dB, and the directivity is greater than 22 dB.

[0049] Uses a separate 100 MHz temperature-controlled crystal oscillator (model MV89A, short-term stability better than 10). -11 The local oscillator (MHz) is used as an independent local oscillator, which is frequency-multiplied to 3.5001 GHz via a phase-locked loop. This local oscillator and the reference source of the RF signal generation unit use independent clock sources and are physically isolated—each is placed in an independent shielded cavity and powered by an independent low-dropout linear regulator to prevent crosstalk. This avoids the impact of common-mode error on frequency offset detection accuracy and ensures the accuracy of frequency detection.

[0050] The coupled signal and the local oscillator signal are mixed by a high-linearity double-balanced mixer ADE-1, and the output is connected to a crystal bandpass filter ECS-100kHz-20k-BP with a center frequency of 100 kHz and a 3 dB bandwidth of 20 kHz. This filter is encapsulated in an independent shielded box, and the input and output impedances are matched to 50 Ω.

[0051] The filtered 100 kHz ± Δf signal is shaped into a square wave by a high-speed comparator TLV3501 (response time 4.5 ns) and sent to the dedicated counting input pin of the field-programmable gate array (FPGA) chip XC7A35T. The FPGA internally implements a frequency counter based on a multi-cycle synchronous measurement method. The counting clock is preferably 250 MHz (generated by a PLL from the system master clock), and the measurement gate time is preferably set to 1 ms, with a theoretical resolution better than 1 Hz.

[0052] 3. Deployment of multi-source sensing units

[0053] The multi-source sensing unit in claim 1 is used to collect multi-dimensional state information related to frequency stability, wherein the multi-dimensional state information includes at least temperature information, power supply voltage information and system runtime information.

[0054] Three DS18B20 digital temperature sensors (accuracy ±0.5°C, resolution 0.0625°C) are mounted in close contact with each other using thermal grease on the surface of the phase-locked loop chip, near the quartz crystal oscillator housing, and near the VCO cavity. They are connected to the GPIO pins of the FPGA via a single bus, and the timing of reading is controlled by the internal state machine of the FPGA. The sampling period is preferably 10 ms.

[0055] An ADS1115 16-bit analog-to-digital converter (ADC) module (sampling rate 10 kHz, I2C bus interface) is connected in parallel to the 3.3 V power supply input of the RF signal generation unit to monitor the power supply voltage in real time. The reference voltage of the ADC is an external high-precision reference source REF3033 (temperature drift 3 ppm / °C).

[0056] A 32-bit counter is maintained inside the field-programmable gate array to record the cumulative running time of the device with an accuracy of 1 ms.

[0057] By distributing temperature sensors in multiple heat-sensitive areas such as the VCO, PLL chip, and crystal oscillator, and combining them with voltage monitoring and runtime statistics, a sufficient data foundation is provided for the subsequent extension of the Kalman filter to separate different drift sources.

[0058] 4. Deployment of Intelligent Decision and Control Unit

[0059] The additional technical features in claims 1, namely, "intelligent decision and control unit, used to receive the frequency offset Δf and the multidimensional state information, fuse the frequency offset Δf and the multidimensional state information based on the extended Kalman filter to separate the frequency offset components caused by temperature drift, aging and power supply disturbance, and calculate the optimal compensation control quantity u based on the model predictive controller", and claims 3, 4 and 5.

[0060] The intelligent decision and control unit is implemented entirely within the field-programmable gate array, occupying approximately 30% of the logic resources (approximately 10,000 lookup tables and 20 DSP48E1 units).

[0061] The extended Kalman filter module uses fixed-point arithmetic (32-bit, 16-bit integer + 16-bit decimal). The state variable dimension is 5 (temperature drift coefficient α, aging drift rate β, voltage sensitivity γ, temperature T, frequency deviation Δf), and the observation variable dimension is 4 (Δf, T, V, t). The update period is 1 ms. The initial state covariance matrix P0 is set as a diagonal matrix with diagonal elements [1e-6, 1e-12, 1e-6, 1e-2, 1e-1]; the process noise covariance Q is set as a diagonal matrix with diagonal elements [1e-8, 1e-14, 1e-8, 1e-4, 1e-2]; and the observation noise covariance R is set as diag([1, 0.1, 1e-4, 1e-6]).

[0062] The model predictive controller module has N = 5 prediction time steps and M = 2 control time steps. The cost function Q... err =100, R u = 1, and the gradient descent method is used to solve the problem. The number of iterations is limited to 10, and the calculation is completed within 1 ms.

[0063] The forgetting factor λ = 0.99 of the recursive least squares module has an update cycle of 1 second, and the model is updated only after the amount of data exceeds 1000 valid sampling points.

[0064] This unit decomposes Δf into three independent drift components in real time through EKF, enabling the system to adopt corresponding compensation strategies for frequency deviations from different sources. Then, it uses MPC for forward prediction and optimal control, which has higher compensation accuracy and faster response speed than traditional PID or lookup table schemes.

[0065] 5. Deployment of Radio Frequency Front-End Execution Unit

[0066] The corresponding technical features are as follows: "RF front-end execution unit, used to perform frequency compensation according to the optimal compensation control quantity u, wherein the RF front-end execution unit includes a fine-tuning path and a coarse-tuning path" in claim 1, and the additional technical features in claims 6 and 7.

[0067] The fine-tuning path uses a 16-bit digital-to-analog converter (DAC) chip, AD5662 (1 MHz update rate). Its output is connected to the VCO tuning voltage input (VTUNE pin) of the phase-locked loop (PLL) chip after passing through a second-order active low-pass filter (10 kHz cutoff frequency). The reference voltage for the DAC is the same REF3033 reference source as the ADC. The DAC output range is 0-3.3 V, corresponding to a VCO tuning range of ±5 kHz, with a resolution of 0.15 Hz / LSB.

[0068] The coarse adjustment path is connected to the register of the phase-locked loop chip via the SPI interface (10 MHz) to modify the PLL division ratio, with the coarse adjustment step set to 100 Hz.

[0069] The switching logic is implemented within the intelligent decision and control unit. When |Δf| > 4.5 kHz (with a set margin of 0.5 kHz), a coarse adjustment is first performed by writing the division ratio correction value via SPI to pull the frequency offset back within ±4.5 kHz. Then, the digital-to-analog converter performs fine-tuning. The switching process must be completed within 1 ms, during which transient jitter in the output frequency is suppressed by a PLL loop filter. Through the hierarchical execution strategy of coarse and fine adjustment, the dual requirements of wide-range frequency offset correction and fine frequency adjustment are met, while reducing the tuning burden on the digital-to-analog converter.

[0070] Example 2: System Workflow

[0071] This embodiment details the workflow of the system during the closed-loop operation phase, corresponding to each step in claim 8, and its control flow is as follows: Figure 2 As shown.

[0072] After the system is powered on, it enters the initialization phase: the intelligent decision and control unit closes the compensation loop and records the Δf detected by the frequency error detection unit as the "zero bias" reference value; the extended Kalman filter initializes the state value; and the model predictive controller sets the initial value of the control quantity u to 0.

[0073] After initialization, the system enters the closed-loop operation phase, repeatedly executing the following steps with a period of 1 ms:

[0074] Step 1: Real-time error detection. The field-programmable gate array counter of the frequency error detection unit counts the 100 kHz ± Δf signal within a 1 ms gate time to obtain the frequency measurement value f. meas Calculate Δf = f meas - 100000, the measurement results are transmitted to the extended Kalman filter module of the intelligent decision and control unit via registers.

[0075] Step 2: Multi-source information acquisition. The multi-source sensing unit reads the temperature values ​​from three DS18B20 sensors via a single bus, takes the average value as the system temperature T, and reads the voltage value V from the ADS1115 sensor via the I2C bus. dd Read t from the internal counter op All data is packaged and stored in the BRAM of the field-programmable gate array for the extended Kalman filter to read.

[0076] Step 3: Intelligent Analysis and Decision Making. The extended Kalman filter module reads Δf, T, V, and t, performs prediction and update steps, and outputs the temperature drift component Δf at the current moment.T , Aging component Δf A Power supply disturbance component Δf P The model predictive controller module predicts the frequency offset trend over multiple time steps based on the estimation results of the extended Kalman filter. It then solves the optimization problem using gradient descent to obtain the optimal control sequence [u0, u1, ..., u4], executing only the first control variable u0. Simultaneously, it compares |Δf| with a 4.5 kHz threshold to determine whether to trigger coarse adjustment.

[0077] Step 4: Accurately perform compensation. In fine-tuning mode, the intelligent decision and control unit writes u0 into the SPI register of the digital-to-analog converter, and the digital-to-analog converter outputs the corresponding analog voltage to the VTUNE pin of the VCO to achieve frequency fine-tuning; in coarse-tuning mode, the intelligent decision and control unit first writes the frequency division ratio correction value to the PLL chip via SPI, and then resumes fine-tuning mode after the PLL relocks.

[0078] Step 5: Closed-loop iteration. The system immediately returns to step 1, and the frequency error detection unit detects the adjusted Δf. If it does not converge, steps 2-5 are repeated.

[0079] Self-learning phase: The system records the current Δf, T, V, t and the corresponding compensation amount u into the BRAM circular buffer of the field-programmable gate array. When the number of records exceeds the threshold, the recursive least squares module starts working, updating the device-specific temperature-frequency offset model parameters (temperature drift coefficient α) and aging-frequency offset model parameters (aging drift rate β) online based on these data. The updated parameters are used in the next initialization of the extended Kalman filter.

[0080] The entire closed-loop process runs continuously with a period of 1 ms to ensure that the shielding signal always maintains a precise lock on the target carrier.

[0081] Example 3: System Operation Results

[0082] This embodiment verifies the performance of the above system in terms of frequency stability, shielding success rate, power consumption and cost, long-term stability, spectrum selectivity and power fluctuation suppression through actual measurement data.

[0083] 1. Frequency stability test

[0084] Temperature cycling tests were conducted in a temperature chamber. Without closed-loop control, the frequency shift was significant; with closed-loop control enabled, the frequency shift was significantly suppressed, and stability was greatly improved. These results demonstrate that, through real-time detection by the frequency error detection unit and closed-loop compensation by the intelligent decision-making and control unit, the system can control the frequency shift to a very small range across the entire temperature spectrum.

[0085] 2. Shielding effectiveness verification

[0086] Actual shielding tests were conducted in the 5G NR n78 band. The shielding effect of a conventional shielding device not employing the scheme described in this embodiment was significantly lower than that of the shielding device using the scheme described in this embodiment. This result indicates that, because the frequency is precisely locked to the target carrier, the shielding signal energy can effectively fall into the synchronization channel of the communication equipment, avoiding being filtered out by the AFC mechanism.

[0087] 3. Power Consumption and Cost Comparison

[0088] The isothermal crystal oscillator / wideband solution not using the approach described in this embodiment has higher power consumption and material costs; the approach in this embodiment uses a common quartz crystal oscillator, which significantly reduces both power consumption and material costs. This result demonstrates that the algorithm-for-hardware approach of this invention significantly reduces device cost and power consumption while maintaining performance.

[0089] 4. Long-term stability testing

[0090] Continuous operation tests show that the cumulative frequency drift is small. After automatic calibration over a wide temperature range, the initial frequency deviation is significantly reduced. This result indicates that the system can automatically track and compensate for frequency drift caused by crystal oscillator aging.

[0091] 5. Spectral Selectivity Test

[0092] Spectrum analyzer tests show that the adjacent-channel radiated power of this embodiment at frequencies deviating from the center frequency is significantly lower than that of a typical broadband scheme. This result indicates that due to the high frequency locking accuracy, the system can employ a narrower transmit bandwidth, thereby significantly reducing interference to adjacent frequency bands.

[0093] 6. Power fluctuation suppression test

[0094] When the supply voltage fluctuates, the solution in this embodiment can quickly suppress the frequency offset within milliseconds. This result shows that the extended Kalman filter can quickly identify and separate the frequency offset component caused by power supply disturbances, giving the system good anti-interference capability against power supply fluctuations.

Claims

1. A frequency adaptive stabilization system for a signal shielding device, characterized in that, include: The radio frequency signal generation unit is used to generate shielded signals, including a crystal reference source, a phase-locked loop, and a voltage-controlled oscillator; The frequency error detection unit is used to detect and quantify the frequency offset Δf of the radio frequency output signal in real time through an independent local oscillator and a digital phase detection architecture. The local oscillator and the radio frequency signal generation unit use independent clock sources and are physically isolated from each other. A multi-source sensing unit is used to collect multi-dimensional state information related to frequency stability. The multi-dimensional state information includes at least temperature information, power supply voltage information, and system runtime information. The intelligent decision and control unit is used to receive the frequency offset Δf and the multi-dimensional state information, fuse the frequency offset Δf and the multi-dimensional state information based on the extended Kalman filter to separate the frequency offset components caused by temperature drift, aging and power supply disturbance, and calculate the optimal compensation control quantity u based on the model predictive controller. The radio frequency front-end execution unit is used to perform frequency compensation according to the optimal compensation control quantity u. The radio frequency front-end execution unit includes a fine-tuning path and a coarse-tuning path. The fine-tuning path outputs an analog voltage to the tuning voltage input terminal of the voltage-controlled oscillator through a digital-to-analog converter to achieve fine-tuning of the frequency. The coarse-tuning path achieves coarse-tuning of the frequency by writing a frequency division ratio correction value to the phase-locked loop. The system performs frequency offset detection, analysis, and compensation cyclically at a preset period.

2. The system according to claim 1, characterized in that, The frequency error detection unit includes: A directional coupler is used to couple a portion of the radio frequency transmission signal from the output of the radio frequency front-end execution unit; A mixer is used to mix the coupled signal with the signal from the independent local oscillator; A bandpass filter is used to extract the difference frequency signal from the mixed signal; A comparator is used to shape the difference frequency signal into a square wave; A counter is used to count the shaped square wave signal to calculate the actual frequency of the difference frequency signal, and then to deduce the real-time frequency deviation Δf of the RF output signal.

3. The system according to claim 1, characterized in that, The intelligent decision and control unit employs an extended Kalman filter, using Δf as the observed quantity and temperature change, runtime, and voltage change as inputs. Based on the physical model Δf = αΔT + βt + γΔV + w, it estimates and separates the temperature drift component Δf online. T , Aging component Δf A and power supply disturbance component Δf P , where α is the temperature drift coefficient, β is the aging drift rate, γ is the voltage sensitivity, and w is the process noise.

4. The system according to claim 3, characterized in that, The intelligent decision and control unit employs a model predictive controller. Based on the estimation results of the extended Kalman filter, it uses the system model Δf(k + 1) = Δf(k) + u(k) + w(k) to predict the frequency offset trend over multiple future control cycles, where k is the discrete time index of the current moment. The prediction is achieved by solving the cost function J = Σ(Q... err ∙Δf(k + i) 2 + R u ∙u(k + i) 2 The optimization problem is to calculate the optimal compensation control quantity u, where i is the step index in the prediction time domain, and Q is the value of the control quantity. err R is the penalty weight for frequency offset error. u Penalty weights are applied to changes in the control quantity.

5. The system according to claim 1, characterized in that, The intelligent decision and control unit uses recursive least squares to update the device-specific temperature-frequency deviation model parameters and aging-frequency deviation model parameters online. The forgetting factor λ of the recursive least squares method is 0.

99.

6. The system according to claim 1, characterized in that, When the frequency offset Δf detected by the RF front-end execution unit exceeds the linear tuning range of the fine-tuning path, the frequency offset is first pulled back into the linear operating range of the fine-tuning path through the coarse-tuning path, and then fine-tuned by the fine-tuning path.

7. The system according to claim 6, characterized in that, The linear tuning range of the fine-tuning path is ±5 kHz, and the switching threshold of the coarse-tuning path is ±4.5 kHz.

8. A control method for a frequency adaptive stabilization system of a signal shielding device, characterized in that, Includes the following steps: Step 1: The radio frequency signal generation unit generates a shielded signal and transmits it through the radio frequency front-end execution unit; Step 2: The frequency error detection unit detects the frequency offset Δf of the RF output signal in real time through an independent local oscillator and a digital phase detector architecture; Step 3: The multi-source sensing unit collects temperature information, power supply voltage information, and system runtime information; Step 4: The intelligent decision and control unit receives the frequency offset Δf and the multi-dimensional state information, fuses multi-source data based on the extended Kalman filter to separate the frequency offset components caused by temperature drift, aging and power supply disturbance, and calculates the optimal compensation control quantity u based on the model predictive controller. Step 5: The RF front-end execution unit performs frequency compensation according to the optimal compensation control quantity u. The frequency compensation includes fine-tuning operation by performing frequency fine-tuning through a digital-to-analog converter and coarse-tuning operation by modifying the frequency division ratio of the phase-locked loop. Step 6: Return to Step 2 to form a closed-loop iteration.

9. The method according to claim 8, characterized in that, In step four, the extended Kalman filter uses Δf as the observed quantity and temperature change, runtime, and voltage change as inputs. Based on the physical model Δf = αΔT + βt + γΔV + w, it estimates and separates the temperature drift component Δf online. T , Aging component Δf A and power supply disturbance component Δf P , where α is the temperature drift coefficient, β is the aging drift rate, γ is the voltage sensitivity, and w is the process noise.

10. The method according to claim 8, characterized in that, In step four, the model predictive controller uses the system model Δf(k + 1) = Δf(k) + u(k) + w(k) based on the estimation results of the extended Kalman filter to predict the frequency offset change trend in multiple future control cycles, where k is the discrete time index of the current moment. The optimal compensation control quantity u is calculated by solving a constrained optimization problem, and the constraint of the control quantity u is |u| < 5 kHz.

Images