A radio frequency control method based on tuning response and harmonic optimization

Abstract

This invention discloses a radio frequency (RF) control method based on tuning response and harmonic optimization, relating to the field of RF technology. The method includes: Step S1: Generating an array configuration word in the target RF system using a multi-capacitor switching array; Step S2: Generating an initial RF signal by driving a switch state switch using the multi-capacitor switching array; Step S3: Iteratively calculating notch filter parameters using an iterative optimization method; Step S4: Configuring the iteratively generated optimal notch parameters to the notch filter to filter and correct the initial RF signal. Compared with existing technologies, this invention utilizes a multi-capacitor switching array to acquire RF signal spectral data, constructs an iterative optimization process, and combines iterative optimization of notch filter parameters with optimal notch parameter correction, thereby reducing the system's residual harmonic measurement value. It has the advantages of effectively reducing residual spurious power and maintaining high signal-to-noise ratio and low error performance, while reducing systemic interference.

Description

Technical Field

[0001] This invention relates to the field of radio frequency technology, and more specifically to a radio frequency control method based on tuning response and harmonic optimization. Background Technology

[0002] In modern radio frequency (RF) control methods, such as software-defined radio, radar, particle accelerators, and high-end test instruments, rapid tuning and band switching of the output RF signal are frequently required. This system typically consists of a digital master control unit and an analog RF front-end. The digital controller calculates the target parameters and configures these parameters into the corresponding RF hardware chip via a digital interface to perform operations such as frequency synthesis, power adjustment, and filter switching. Existing RF control methods generally exhibit a tendency to prioritize steady-state performance over transient performance in their design and data processing. The system's control algorithms and data processing units focus solely on managing and optimizing parameters in a steady state, easily neglecting the fine-grained management of waveforms and data during switching.

[0003] In existing RF systems, when performing continuous and rapid tuning of RF signals, digital controllers typically employ a simple and crude step control strategy after a band or frequency switching command is issued. For example, to change the amplifier's bias point, the controller directly writes the new value into the digital-to-analog converter register, causing a momentary jump in the amplifier's bias voltage. Similarly, the frequency synthesizer's configuration parameters are also updated instantaneously to quickly establish phase lock at the new frequency. This lack of transient process management can lead to pulse step responses in the frequency domain of the generated RF signal, which may contain many high-frequency components. These pulse step responses generate a large number of wideband harmonics and spurious spectra. These transient harmonic noises severely degrade the spectral quality during switching. Furthermore, because these pulse step responses are generated instantaneously during switching, they are difficult to effectively filter out by steady-state filters, thus directly increasing the overall harmonic residual measurement value of the system and affecting the purity of the RF signal. Summary of the Invention

[0004] This invention provides a radio frequency (RF) control method based on tuning response and harmonic optimization, which solves the problem that existing RF systems lack effective quality correction for RF signals that contain many high-frequency pulse responses due to frequency switching-induced bias voltage jumps during continuous and rapid tuning.

[0005] This invention is achieved through the following technical solution:

[0006] A radio frequency control method based on tuning response and harmonic optimization, the method comprising:

[0007] Step S1: Preset a digital controller and a notch filter in the target RF system, preset a multi-capacitor switching array in the digital controller, execute the target RF system to process the target signal source, use the digital controller to receive the frequency switching command from the target RF system, and use the multi-capacitor switching array to generate the array configuration word corresponding to the command.

[0008] Step S2: Based on the high-speed parallel interface, the array configuration word is written into the drive register of the multi-capacitor switching array. The drive register is used to drive the switching state of the multi-capacitor switching array to complete the switching within a nanosecond time period. After the switching is completed, the initial radio frequency signal is generated, and the spectrum data of the initial radio frequency signal in the frequency domain is obtained.

[0009] Step S3: Based on the spectrum data, extract the initial spurious energy parameters from the initial radio frequency signal, construct an iterative optimization method, use the notch filter parameters as input data for the iterative optimization method, use the initial spurious energy parameters as constraint variables for the iterative optimization method to perform iterative calculation, and label the calculation results as the optimal notch parameters;

[0010] Step S4: Configure the optimal notch filter parameters as configuration parameters to the notch filter, use the notch filter to filter the initial RF signal and mark it as the corrected RF signal, and verify the parameters. When the corrected RF signal meets the verification, it is output through the target RF system. When the corrected RF signal does not meet the verification, return to step S3.

[0011] In existing radio frequency (RF) systems, during continuous and rapid tuning of RF signals, the amplifier's bias voltage undergoes a momentary jump upon receiving a band or frequency switching command. This can cause the generated RF signal to exhibit a pulse step response containing numerous high-frequency components in the frequency domain. This pulse step response generates a large amount of wideband harmonic and spurious spectrum. These transient harmonic noises severely degrade the spectral quality during the switching process. Furthermore, because these pulse step responses occur instantaneously during switching, they are difficult to effectively filter out using steady-state filters, directly increasing the overall harmonic residual measurement value of the system. Therefore, this invention provides an RF control method based on tuning response and harmonic optimization to address the lack of effective quality correction for RF signals with numerous high-frequency pulse responses due to frequency switching-induced bias voltage jumps during continuous and rapid tuning in existing RF systems.

[0012] Furthermore, the iterative optimization method is constructed using a genetic algorithm, the process of which includes:

[0013] Step A1: Preset the number of iterations, encode the parameters in the notch filter parameters into individuals representing genetic genes in vector form; randomly initialize and generate several candidate individuals, each individual representing a set of notch filter configuration parameters;

[0014] Step A2: Construct a fitness function using the initial stray energy parameter as a constraint parameter, perform quality screening on all individuals, iterate through individuals that meet the quality screening according to the normal fitness value, and remove the remaining individuals;

[0015] Step A3: In each iteration, select two groups of individuals as parents based on their fitness values ​​and perform partial parameter swapping to generate new individuals representing offspring. After reaching the maximum number of iterations, output the generated individuals as the optimal notch filter parameters.

[0016] Furthermore, the fitness function is set to the following form:

[0017] Let the initial spurious energy parameter in the initial RF signal spectrum data be denoted as Es, the maximum spurious peak value of the initial RF signal band be denoted as Pe, and the delay insertion loss value of the notch filter be denoted as De; let the fitness function be denoted as Q, and let the individual be represented by the independent variable k.

[0018] The fitness function is then expressed as: Q(k) = ω1∙Es + ω2∙Pe(k) + ω2∙De(k).

[0019] Where ω1, ω2, and ω3 represent the weighting coefficients of the initial stray energy parameter Es, the maximum stray peak value Pe, and the delay insertion loss value De, respectively.

[0020] Furthermore, the method for quality screening of all individuals is set as follows:

[0021] Set the system frequency range of the target RF system to the range of notch frequency values ​​in the individual's parameter vector, and set the parameter range for the quality factor in the individual's parameter vector; when the individual's notch frequency is within the system frequency range and the quality factor is within the parameter range, the current individual is judged to meet the quality screening; when the notch frequency exceeds the system frequency range or the quality factor exceeds the parameter range, the current individual is judged to not meet the quality screening.

[0022] Furthermore, redundant bits are added to the array configuration word, and a synchronous clock is set to latch the configuration word; an adjustable phase synchronous clock is assigned to the different weight capacitors in the multi-capacitor switching array, and CRC check is performed before data is written to the latch unit to synchronously regenerate erroneous data.

[0023] Furthermore, the parameter verification process includes:

[0024] Based on the historical execution records of the target RF system, historical spurious energy parameters generated each time an RF signal is generated are collected, and an energy threshold is set for all the historical spurious energy parameters. The corrected spurious energy parameters of the corrected RF signal are extracted. When the corrected spurious energy parameters are lower than the energy threshold, the corrected RF signal is determined to be compliant with verification. When the corrected spurious energy parameters are higher than the energy threshold, the corrected RF signal is determined to be non-compliant with verification.

[0025] Furthermore, the parameter samples of historical stray energy parameters undergo precision optimization processing, the process of which includes:

[0026] Step M1: Preprocess the collected historical spurious energy parameters, including removing invalid values, processing outliers and filling missing values, normalizing them based on the same scale, and representing the single execution process of the target radio frequency system in periodic form;

[0027] Step M2: Set the historical stray energy parameters of the most recent few periods as target sample points, and set the historical stray energy parameters of the remaining periods as regular sample points. Calculate the data distance value from each target sample point to each regular sample point.

[0028] Step M3: Sort all data distance values ​​according to size, set a percentage threshold for data distance values ​​from low to high, and collect and label the data distance values ​​within the percentage threshold as qualified distance values ​​after rounding to the nearest integer.

[0029] Step M4: Compare the data correlation between each of the selected regular sample points with each target sample point, and replace each target sample point with the regular sample point with the highest data correlation.

[0030] Furthermore, let the results of normalizing each feature individual using zero mean for the target sample points and regular sample points be the first distribution values ​​E. ij1 Second distribution value F ij2 And if the data distance value is set to D, then the formula for calculating the data distance value D is: Where i represents the feature ordinal number, X represents the target sample point, Y represents the regular sample point, the number of features in each historical stray energy parameter is set to n, j1 represents the target sample point ordinal number, and j2 represents the regular sample point ordinal number.

[0031] Furthermore, the calculation process of the data distance value includes: setting the total number of periods up to the current time as N, and setting the feature individual representation of the target sample point as X. ij Where i represents the feature ordinal number, and j represents the sample ordinal number of the current sample point within the corresponding category of sample points; let the mean and standard deviation of each feature individual within each sample point be represented by μ. i and σi Then μ i and σ i The calculation formulas are expressed as follows:

[0032] The mean μ of each individual characteristic i The calculation formula is expressed as: ,

[0033] The standard deviation σ of each characteristic individual i The calculation formula is expressed as: ;

[0034] For each current period, all characteristic individuals within the historical stray energy parameters are normalized using zero-mean normalization. Let the total distribution value of the processed result be represented by Z. ij Its calculation formula is expressed as: ,

[0035] Where μ ij σ represents the mean of the i-th feature individual in the j-th sample within the corresponding category sample points. ij This represents the standard deviation of the i-th feature of the j-th sample within the corresponding category of sample points;

[0036] Let the first distribution value E ij1 Second distribution value F ij2 The calculation process all conforms to the total distribution value Z. ij The calculation formula.

[0037] Furthermore, the data correlation between the target sample points and regular sample points is calculated based on cosine similarity, including:

[0038] Let Y represent the characteristic individuals of the regular sample points. ij Let the number of features within each historical stray energy parameter be n, and let the similarity value be C. Then the vector dot product is expressed as: The vector norm of the features within the target sample points is expressed as: The vector norm of features within a regular sample point is expressed as: Then the formula for calculating the similarity value C is: .

[0039] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0040] 1. By acquiring the spectrum data of the radio frequency signal immediately after the frequency switching is completed and extracting the spurious energy parameters, transient harmonics and spurious signals are identified in real time. Nanosecond-level fast frequency switching is achieved with the help of a multi-capacitor switching array, avoiding frequency switching lag caused by the step-by-step adjustment process of the traditional digital controller.

[0041] 2. By constructing an iterative optimization process and combining iterative optimization of notch filter parameters, the notch filter parameters can be continuously adjusted according to the real-time spectrum results. While ensuring fast switching, the broadband spurious signals caused by the impulse response are effectively suppressed, improving the ability to suppress harmonics under complex operating conditions, and achieving fast switching and clean output.

[0042] 3. The system uses optimal notch parameters to correct the large number of high-frequency components and broadband spurious signals generated during the switching instant, effectively reducing residual spurious power and reducing the pollution of the spectrum by the pulse step response. This reduces the measured value of system harmonic residuals, improves the purity of the radio frequency signal, and maintains high signal-to-noise ratio and low error performance, thereby reducing systemic interference. Attached Figure Description

[0043] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0044] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0046] Example 1

[0047] like Figure 1 As shown, this invention provides a radio frequency control method based on tuning response and harmonic optimization, the method comprising:

[0048] Step S1: Preset a digital controller and a notch filter in the target RF system, preset a multi-capacitor switching array in the digital controller, execute the target RF system to process the target signal source, use the digital controller to receive the frequency switching command from the target RF system, and use the multi-capacitor switching array to generate the array configuration word corresponding to the command.

[0049] Step S2: Based on the high-speed parallel interface, the array configuration word is written into the drive register of the multi-capacitor switching array. The drive register is used to drive the switching state of the multi-capacitor switching array to complete the switching within a nanosecond time period. After the switching is completed, the initial radio frequency signal is generated, and the spectrum data of the initial radio frequency signal in the frequency domain is obtained.

[0050] Step S3: Based on the spectrum data, extract the initial spurious energy parameters from the initial radio frequency signal, construct an iterative optimization method, use the notch filter parameters as input data for the iterative optimization method, use the initial spurious energy parameters as constraint variables for the iterative optimization method to perform iterative calculation, and label the calculation results as the optimal notch parameters;

[0051] Step S4: Configure the optimal notch filter parameters as configuration parameters to the notch filter, use the notch filter to filter the initial RF signal and mark it as the corrected RF signal, and verify the parameters. When the corrected RF signal meets the verification, it is output through the target RF system. When the corrected RF signal does not meet the verification, return to step S3.

[0052] The digital controller, located at the core of the system's control and data processing, performs frequency tuning, parameter calculation, command distribution, and real-time monitoring. In practical applications, it typically has a built-in high-speed interface for receiving switching commands and configuration parameters, and pre-stores capacitor array configuration tables for different frequency points. This ensures that upon receiving a switching command, it can quickly issue instructions and drive the RF front-end to perform operations. The notch filter, as a key filtering structure in the RF signal chain, selectively suppresses spurious components or harmonics in specific frequency bands, thereby ensuring the spectral purity of the output signal. The digital controller's pre-defined multi-capacitor switching array means that its software / firmware pre-designs and stores capacitor array configuration mapping tables corresponding to different frequency switching commands. In practical applications, this can be achieved by setting a frequency point and a configuration word one-to-one correspondence, allowing the digital controller to directly access the capacitor drive switches in the multi-capacitor switching array through a high-speed parallel interface. The digital controller not only receives frequency switching commands but also calls the array configuration word corresponding to the command based on a built-in algorithm. These configuration words determine which capacitors in the multi-capacitor array are connected and which are disconnected. During implementation, a logic module can be pre-written using a hardware language to generate control signals for the multi-capacitor array. This module is then bound to the register address space of the digital controller, which only needs to write a configuration word to complete the array switching.

[0053] The term "data channel" refers to the data channel between the digital controller and the multi-capacitor switching array. It features high bit width, low latency, and the ability to transmit multiple bits simultaneously, typically utilizing an LVDS bus or parallel GPIO bus. The array configuration word is a set of binary control codes generated and output by the digital controller, used to indicate the on or off state of each capacitor switch in the multi-capacitor switching array; essentially, it is a digital bit mode, corresponding to the operating state of each switching unit in the array. The digital controller first generates the array configuration word and writes it to the driver register via a high-speed parallel interface. The driver register acts as a temporary storage and distribution unit, latching the configuration word at the clock edge and simultaneously distributing control signals to the control terminals of each switch in the multi-capacitor switching array. The multi-capacitor switching array typically consists of multiple MOSFET switching devices and capacitors of varying capacitances. When the output level of the driver register changes, all switching devices synchronously toggle their states, thereby rapidly changing the equivalent capacitance value of the array and reconfiguring the capacitor combination. Because the drive link from register to switch is short and highly parallel, signal delay and jitter are minimal. The equivalent capacitance switching of the entire array can be completed on the order of nanoseconds, far faster than the microsecond or even millisecond-level adjustment of traditional DAC step-mode. This approach enables the RF front-end to complete tuning or bias point switching in a very short time, reducing transient distortion caused by delayed or inconsistent switching; at the same time, it ensures that modules such as frequency synthesizers, amplifier biasing, or filtering networks can immediately enter the new operating state.

[0054] The initial RF signal represents the RF signal generated after a faster bias switching. While it has undergone preliminary correction, it may still contain high-frequency pulse step responses and broadband spurious spectra at the moment of switching, requiring further filtering. The initial RF signal is converted from the time domain to the frequency domain, and the amplitude and phase information of the signal at each frequency component is obtained through Fourier transform or digital spectrum analysis. The spectrum data typically includes the amplitude, frequency value, and phase information at each frequency point, and spurious energy, harmonic content, and broadband noise level can be quantitatively extracted from the frequency domain data. Sampling methods include using a high-speed analog-to-digital converter to sample the RF signal into the digital domain, followed by Fourier transform to calculate the spectrum; or using a spectrum analyzer or RF power meter to scan a specific frequency band to obtain amplitude-frequency information.

[0055] The initial spurious energy parameter can represent a single parameter or a set of parameters in specific implementations; when represented as a set of parameters, it is represented by a vector. The initial spurious energy parameter represents the non-fundamental frequency components in the initial radio frequency signal, such as higher harmonics, broadband impulse responses, and non-ideal frequency energy generated by impulse step responses. The process of extracting spurious energy parameters from the radio frequency signal can be implemented by selecting a frequency range outside the target frequency band, integrating the power spectrum of that range to obtain the spurious energy, and further weighting different frequency bands to obtain the spurious energy parameter. In the instrument, a spectrum analyzer can be used to directly provide out-of-band spurious power measurement. This method is applicable to the extraction and recording of initial and historical spurious energy parameters. The iterative optimization method enables the optimal parameters of the notch filter to be continuously adjusted according to the measurement results, ensuring that the notch filter parameters can adapt to the transient spurious signals of the radio frequency signal, rather than being fixed. In specific implementations, optimization algorithms such as gradient descent, conjugate gradient, and least mean square iteration can be used to find the optimal parameters. The initial spurious energy parameter is introduced into the optimization algorithm as an evaluation metric. During the iteration process, the notch filter parameter selection must satisfy the constraint of reducing spurious energy. For example, if the initial spurious energy is -30 dBc, the algorithm goal might be to iteratively optimize until the spurious energy is reduced to below -60 dBc. The iterative optimization method continuously adjusts the notch filter parameters until a set of parameters is found that minimizes spurious energy or satisfies the threshold condition. This set of parameters is then applied to the system as the optimal notch filter parameters.

[0056] Example 2

[0057] In this embodiment, the iterative optimization method is constructed using a genetic algorithm, and the process includes:

[0058] Step A1: Preset the number of iterations, encode the parameters in the notch filter parameters into individuals representing genetic genes in vector form; randomly initialize and generate several candidate individuals, each individual representing a set of notch filter configuration parameters;

[0059] Step A2: Construct a fitness function using the initial stray energy parameter as a constraint parameter, perform quality screening on all individuals, iterate through individuals that meet the quality screening according to the normal fitness value, and remove the remaining individuals;

[0060] Step A3: In each iteration, select two groups of individuals as parents based on their fitness values ​​and perform partial parameter swapping to generate new individuals representing offspring. After reaching the maximum number of iterations, output the generated individuals as the optimal notch filter parameters.

[0061] The notch parameters refer to the adjustable parameters of the notch filter, such as the notch center frequency, bandwidth, and depth attenuation. Vector encoding represents combining these parameters into an ordered vector. Each individual in the genetic algorithm represents a candidate filter configuration, and the algorithm selects and evolves among multiple individuals. Random initialization means that within the allowed range of parameters (e.g., center frequency between 1 GHz and 5 GHz, bandwidth between 1 MHz and 50 MHz), several sets of parameter vectors are randomly generated as the initial population. Each individual represents a set of configuration parameters; that is, different individuals correspond to different notch filter configurations. For example, individual 1 corresponds to a narrow bandwidth notch at 2.4 GHz, and individual 2 corresponds to a wide bandwidth notch at 3.1 GHz. In the genetic algorithm, the fitness function determines the quality of candidate individuals in the optimization problem. If the candidate notch parameters cannot reduce the spurious energy to within the threshold, the fitness function will impose a numerical penalty on the solution, forcing the algorithm to move away from these solutions. Through continuous iteration, the fitness function is used to ensure that the notch parameters selected by the genetic algorithm can reduce spurious energy while preserving the effective components of the RF signal to the greatest extent, thereby outputting a high-quality calibrated RF signal.

[0062] Each individual represents a set of notch filter configuration parameters. During iteration, the genetic algorithm prioritizes individuals with higher fitness values ​​as parents to ensure that superior genes have a greater chance of being passed on. The parameter vectors of the parent individuals undergo "partial exchange," simulating gene crossover in genetics. For example, if parent 1 vector = [f1, Q1] and parent 2 vector = [f2, Q2], crossover might result in offspring 1 = [f1, Q2] and offspring 2 = [f2, Q1], allowing new individuals to inherit the superior characteristics of their parents while generating new combinations and increasing search diversity. Each crossover operation produces new individuals representing new notch filter parameter configurations. These offspring will continue to participate in fitness evaluation and selection as candidate solutions in the next iteration. After multiple iterations, individuals with lower fitness are gradually eliminated, while individuals with higher fitness continue to propagate. After the number of iterations reaches a preset upper limit, the algorithm selects the optimal individual from the final generation as the final optimal notch filter parameters. By utilizing the selection and crossover mechanism of a genetic algorithm, high-fitness individuals are allowed to reproduce as parents. Through multiple rounds of iteration, the parameter combination of the notch filter is continuously optimized, eventually automatically converging to the optimal notch parameters to achieve the best suppression of RF signal spurious components.

[0063] The spurious energy parameter reflects the initial spurious energy parameter in the radio frequency signal after frequency switching. If the notch filter parameter is not configured properly, the filtered signal may still contain too much spurious energy, thus affecting the quality of the radio frequency signal. Therefore, the spurious energy parameter is used as a threshold constraint to determine whether the configuration parameter combination of an individual is feasible. As a feasible implementation method, in specific applications, the method for quality screening of all individuals is set as follows: the system frequency range of the target radio frequency system is set as the value range of the notch frequency in the individual's parameter vector, and a parameter range is set for the quality factor in the individual's parameter vector; when the individual's notch frequency is within the system frequency range and the quality factor is within the parameter range, the current individual is judged to meet the quality screening; when the notch frequency exceeds the system frequency range, or the quality factor exceeds the parameter range, the current individual is judged to not meet the quality screening. The operating frequency range of the target radio frequency system is directly used as the feasible value range of the notch frequency in the individual's chromosome. For example, if the system operates at 2.5GHz~3.5GHz, the notch frequency parameter in the genetic algorithm individual encoding can only take values ​​within this range; those exceeding this range are considered invalid individuals. Setting a parameter range for the quality factor can avoid both excessively wide bandwidth leading to shallow notches and excessively narrow bandwidth causing reduced implementation difficulty and stability. The implementation method, through the dual constraints of frequency and quality factor, ensures that the notch filter parameters generated by the genetic algorithm do not deviate from the physical implementation conditions of the system, avoiding the algorithm finding a solution that is "mathematically optimal but not practically usable." Locking the notch frequency within the system's operating range ensures that the optimized parameters can effectively suppress spectral spurious signals, while limiting the quality factor within a controllable range prevents the filter from being overly sensitive or too lenient, improving long-term operational stability.

[0064] Furthermore, as a feasible implementation, the fitness function is set to the following form:

[0065] Let the initial spurious energy parameter in the initial RF signal spectrum data be denoted as Es, the maximum spurious peak value of the initial RF signal band be denoted as Pe, and the delay insertion loss value of the notch filter be denoted as De; let the fitness function be denoted as Q, and let the individual be represented by the independent variable k.

[0066] The fitness function is then expressed as: Q(k) = ω1∙Es + ω2∙Pe(k) + ω2∙De(k).

[0067] Where ω1, ω2, and ω3 represent the weighting coefficients of the initial stray energy parameter Es, the maximum stray peak value Pe, and the delay insertion loss value De, respectively.

[0068] The maximum spurious peak value represents the amplitude of the largest single-point spurious component within the operating frequency band of the RF signal. High peak values ​​can cause local interference to downstream receivers or measurement equipment. The delay insertion loss value represents the delay and insertion loss introduced by the notch filter within the signal bandwidth. Excessive delay insertion loss may reduce signal amplitude or affect phase characteristics, thus reducing system performance. The initial spurious energy parameter is added to the maximum spurious peak value to reflect the spurious distribution and peak situation in the signal spectrum, ensuring that the optimization process focuses not only on the total energy but also on local peaks. Pe(k) and De(k) represent the maximum spurious peak value and delay insertion loss value in the configuration parameters represented by the k-th individual. Q(k) represents the fitness value of the k-th individual, indicating the degree of matching between the k-th individual and the target performance requirements; a larger Q(k) value indicates a stronger correlation between the target sample point and the best-matching conventional sample point. The delay insertion loss value is used to control the delay and insertion loss introduced by the filter, avoiding damage to the signal bandwidth or phase characteristics while suppressing spurious signals. By adjusting ω1, ω2, and ω3, optimization can be flexibly performed according to actual system requirements. If high requirements are needed for spurious suppression, the weight values ​​of ω1 and ω2 can be increased; if high requirements are needed for signal amplitude and phase characteristics, the weight value of ω3 can be increased.

[0069] Furthermore, as a feasible implementation, redundant bits are added to the array configuration word, and a synchronous clock is set to latch the configuration word; an adjustable phase synchronous clock is assigned to the different weight capacitors in the multi-capacitor switching array, and CRC check is performed before data is written to the latch unit to synchronously regenerate erroneous data.

[0070] Redundant bits are used to store verification information and error correction codes. During data transmission or writing to the latch, individual bit errors can be detected and corrected through redundant bits, improving data reliability. Before being written to the multi-capacitor switching array, the configuration word is latched by a synchronous clock. This latching ensures strict timing alignment for each configuration switch, preventing switch state misalignment caused by asynchronous writing by the controller. For a multi-capacitor array, each capacitor has a different weight, and their amplitude and phase contributions to the RF output also differ. By allocating synchronous clocks with different phases, the timing of each capacitor switch can be precisely controlled at the nanosecond level, reducing transient interference and step response amplitude. Essentially, this smooths the step response and improves the spectral purity of the switching transient. The CRC is an error detection method based on binary polynomial division, used to detect whether errors have occurred in digital data during transmission or storage. By adding redundant bits to the array configuration word, using synchronous clock latching, allocating adjustable phase clocks, and performing CRC verification and error regeneration, the reliability, amplitude and phase accuracy, and transient spurious suppression of the multi-capacitor switching array in high-speed switching are achieved, effectively improving the spectral quality of the RF signal and system stability.

[0071] Example 3

[0072] In this embodiment, the parameter verification process includes:

[0073] Based on the historical execution records of the target RF system, historical spurious energy parameters generated each time an RF signal is generated are collected, and an energy threshold is set for all the historical spurious energy parameters. The corrected spurious energy parameters of the corrected RF signal are extracted. When the corrected spurious energy parameters are lower than the energy threshold, the corrected RF signal is determined to be compliant with verification. When the corrected spurious energy parameters are higher than the energy threshold, the corrected RF signal is determined to be non-compliant with verification.

[0074] Historical spurious energy parameters generated during previous signal generation processes of the target RF system are collected. By analyzing this historical data, an energy threshold is set to provide a quantitative standard for evaluating the correction signal. This energy threshold can be set based on empirical rules or the mean. If the correction spurious energy is below the energy threshold, the corrected RF signal is deemed to meet the requirements; if the correction spurious energy is above the energy threshold, the corrected RF signal is deemed to not meet the requirements and optimization needs to be re-executed. This historical data-driven energy threshold determination enables RF signal correction quality verification to possess adaptability, closed-loop reliability, and a quantitative standard. It effectively ensures that the corrected RF signal maintains high purity during frequency switching and supports iterative closed-loop control for optimization.

[0075] Furthermore, as a feasible implementation method, the parameter samples of historical stray energy parameters are subjected to precision optimization processing, the process of which includes:

[0076] Step M1: Preprocess the collected historical spurious energy parameters, including removing invalid values, processing outliers and filling missing values, normalizing them based on the same scale, and representing the single execution process of the target radio frequency system in periodic form;

[0077] Step M2: Set the historical stray energy parameters of the most recent few periods as target sample points, and set the historical stray energy parameters of the remaining periods as regular sample points. Calculate the data distance value from each target sample point to each regular sample point.

[0078] Step M3: Sort all data distance values ​​according to size, set a percentage threshold for data distance values ​​from low to high, and collect and label the data distance values ​​within the percentage threshold as qualified distance values ​​after rounding to the nearest integer.

[0079] Step M4: Compare the data correlation between each of the selected regular sample points with each target sample point, and replace each target sample point with the regular sample point with the highest data correlation.

[0080] The processes of removing invalid values, handling outliers, and filling missing values ​​ensure the integrity and cleanliness of historical spurious energy parameter data, reducing interference from anomalous data in subsequent analysis. Normalization unifies spurious energy parameters from different periods or frequencies to the same scale, facilitating subsequent distance calculations and comparisons. Each RF signal generation process is represented by a time period, forming a structured sample sequence. The target sample point represents historical spurious energy parameters from the most recent few periods as the primary focus, used to assess the current system state. The regular sample point represents data from other periods used as reference samples for similarity analysis with the target sample. The data distance calculation involves calculating the data distance between each target sample point and all regular sample points to quantify sample similarity.

[0081] Sort all data distance values ​​from smallest to largest and set a percentage threshold. This percentage threshold can be in the form of a minimum distance, for example, the distances in the top 30% can be set as the minimum. In specific applications, this can be set according to empirical rules of data similarity. Select regular sample points corresponding to distances within the percentage threshold and label them as qualified distance values ​​as reliable reference data. For each target sample point, calculate its correlation with qualified regular sample points, and replace the target sample point with the regular sample point with the highest correlation, thus optimizing accuracy and making historical data more accurate. By performing anomaly removal, normalization, distance filtering, and correlation optimization on historical spurious energy parameters, historical data becomes more accurate and reliable, providing a solid data foundation for RF signal parameter verification and closed-loop optimization, thereby improving the performance of the entire RF control system in spurious suppression and signal purity during rapid frequency switching.

[0082] Furthermore, as a feasible implementation method, let the results of normalizing each feature individual using zero mean for the target sample points and regular sample points be the first distribution value E. ij1 Second distribution value F ij2 And if the data distance value is set to D, then the formula for calculating the data distance value D is: Where i represents the feature ordinal number, X represents the target sample point, Y represents the regular sample point, the number of features in each historical stray energy parameter is set to n, j1 represents the target sample point ordinal number, and j2 represents the regular sample point ordinal number.

[0083] The data distance value represents a similarity measure between the target sample point X and the regular sample point Y. A smaller value indicates greater similarity; a larger value indicates greater difference. The target sample point represents the currently selected historical stray energy data sample that needs optimization analysis; the regular sample point represents the remaining unselected historical stray energy data samples used as references. The number of features represents the number of feature dimensions contained in each sample point, such as stray energy values ​​at different frequencies in the frequency domain, or energy values ​​at different sampling times. The normalization process is used to eliminate dimensional differences, ensuring a balanced contribution of each feature to the distance calculation. (E) ij1 -F ij2 The ) part represents the feature difference, that is, the difference between the target sample and the reference sample in each dimension. The (E) ij1 -F ij2 ) 2 The variance is guaranteed to be non-negative, while highlighting the contribution of larger variances. The first distribution value E ij1 E represents the normalized value of the ith feature in the j1-th target sample point, and the second distribution value is E. ij2 This represents the normalized value of the i-th feature in the j2-th regular sample point. The summation and square root of the differences between the target sample and the regular samples across all feature dimensions yield the true distance, which serves as the final measure of the overall similarity between the target sample point and the regular sample points.

[0084] Furthermore, as a feasible implementation method, the calculation process of the data distance value includes: setting the total number of periods up to the current time as N, and setting the feature individual representation of the target sample point as X. ij Where i represents the feature ordinal number, and j represents the sample ordinal number of the current sample point within the corresponding category of sample points; let the mean and standard deviation of each feature individual within each sample point be represented by μ. i and σ i Then μ i and σ i The calculation formulas are expressed as follows:

[0085] The mean μ of each individual characteristic i The calculation formula is expressed as: ,

[0086] The standard deviation σ of each characteristic individual i The calculation formula is expressed as: ;

[0087] For each current period, all characteristic individuals within the historical stray energy parameters are normalized using zero-mean normalization. Let the total distribution value of the processed result be represented by Z. ij Its calculation formula is expressed as: ,

[0088] Where μij σ represents the mean of the i-th feature individual in the j-th sample within the corresponding category sample points. ij This represents the standard deviation of the i-th feature of the j-th sample within the corresponding category of sample points;

[0089] Let the first distribution value E ij1 Second distribution value F ij2 The calculation process all conforms to the total distribution value Z. ij The calculation formula.

[0090] Because the first distribution value E ij1 Second distribution value F ij2 The calculation formulas are the same, therefore the ordinal j is used to uniformly represent the target sample point ordinal j1 and the regular sample point ordinal j2 in the common calculation formula. The μ ij μ represents the average value of the i-th feature of the j-th sample over all periods, used to reflect the central trend of historical samples and to eliminate the offset of feature values. ij Indicates the σ ij X represents the dispersion of the i-th feature of the j-th sample across all periods, used to measure the magnitude of data fluctuation and to normalize the feature scale. ij This represents the original feature individuals in the target sample points, specifically the historical stray energy value of the i-th feature and the j-th sample. (X) ij - μ i The zero-mean normalization process involves subtracting the mean from the sample features, centering the data around zero and eliminating feature bias. Dividing the zero-mean normalized data by the standard deviation standardizes the numerical fluctuation range of different features, facilitating subsequent similarity calculations. By statistically analyzing the mean and variance of historical spurious energy parameters and calculating the standardized features of the target and regular samples based on zero-mean normalization, the accuracy and robustness of subsequent Euclidean distance calculations are ensured, providing a reliable data foundation for historical sample optimization, spurious energy threshold setting, and closed-loop RF signal correction.

[0091] Furthermore, as a feasible implementation method, the data correlation between the target sample points and regular sample points is calculated based on cosine similarity, which includes:

[0092] Let Y represent the characteristic individuals of the regular sample points. ij Let the number of features within each historical stray energy parameter be n, and let the similarity value be C. Then the vector dot product is expressed as: The vector norm of the features within the target sample points is expressed as: The vector norm of features within a regular sample point is expressed as: Then the formula for calculating the similarity value C is: .

[0093] The vector inner product represents the overall matching degree between the target sample and the regular sample on various features. The norm of the target vector and the norm of the regular vector are used to normalize the inner product. Cosine similarity calculation evaluates the directional similarity between the two vectors. The closer the calculated value is to 1, the more consistent the vector directions, indicating a high degree of similarity in feature distribution. Calculating directional similarity, rather than absolute numerical differences, using the target sample point and the regular sample point as vectors emphasizes overall distribution consistency. In optimizing the accuracy of historical spurious energy parameters, replacing the target sample point with the regular sample point with the one having the highest cosine similarity maximizes the preservation of data feature consistency. By calculating the cosine similarity between the target sample point and the regular sample point, the consistency of the two samples in the direction of feature distribution can be quantified, providing a reliable similarity measure for the accuracy optimization and closed-loop correction of historical spurious energy parameters, achieving high-precision transient quality management of RF signals.

[0094] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A radio frequency control method based on tuning response and harmonic optimization, characterized in that, The method includes: Step S1: Preset a digital controller and a notch filter in the target RF system, preset a multi-capacitor switching array in the digital controller, execute the target RF system to process the target signal source, use the digital controller to receive the frequency switching command from the target RF system, and use the multi-capacitor switching array to generate the array configuration word corresponding to the command. Step S2: Based on the high-speed parallel interface, the array configuration word is written into the drive register of the multi-capacitor switching array. The drive register is used to drive the switching state of the multi-capacitor switching array to complete the switching within a nanosecond time period. After the switching is completed, the initial radio frequency signal is generated, and the spectrum data of the initial radio frequency signal in the frequency domain is obtained. Step S3: Based on the spectrum data, extract the initial spurious energy parameters from the initial radio frequency signal, construct an iterative optimization method, use the notch filter parameters as input data for the iterative optimization method, use the initial spurious energy parameters as constraint variables for the iterative optimization method to perform iterative calculation, and label the calculation results as the optimal notch parameters; Step S4: Configure the optimal notch filter parameters as configuration parameters to the notch filter, use the notch filter to filter the initial RF signal and mark it as the corrected RF signal, and verify the parameters. When the corrected RF signal meets the verification, it is output through the target RF system. When the corrected RF signal does not meet the verification, return to step S3.

2. The radio frequency control method based on tuning response and harmonic optimization according to claim 1, characterized in that, The process of constructing the iterative optimization method using a genetic algorithm includes: Step A1: Preset the number of iterations, encode the parameters in the notch filter parameters into individuals representing genetic genes in vector form; randomly initialize and generate several candidate individuals, each individual representing a set of notch filter configuration parameters; Step A2: Construct a fitness function using the initial stray energy parameter as a constraint parameter, perform quality screening on all individuals, iterate through individuals that meet the quality screening according to the normal fitness value, and remove the remaining individuals; Step A3: In each iteration, select two groups of individuals as parents based on their fitness values ​​and perform partial parameter swapping to generate new individuals representing offspring. After reaching the maximum number of iterations, output the generated individuals as the optimal notch filter parameters.

3. The radio frequency control method based on tuning response and harmonic optimization according to claim 2, characterized in that, The fitness function is set to the following form: Let the initial spurious energy parameter in the initial RF signal spectrum data be denoted as Es, the maximum spurious peak value of the initial RF signal band be denoted as Pe, and the delay insertion loss value of the notch filter be denoted as De; let the fitness function be denoted as Q, and let the individual be represented by the independent variable k. The fitness function is then expressed as: Q(k) = ω1∙Es + ω2∙Pe(k) + ω2∙De(k). Where ω1, ω2, and ω3 represent the weighting coefficients of the initial stray energy parameter Es, the maximum stray peak value Pe, and the delay insertion loss value De, respectively.

4. The radio frequency control method based on tuning response and harmonic optimization according to claim 2, characterized in that, The method for quality screening of all individuals is set as follows: Set the system frequency range of the target RF system to the range of notch frequency values ​​in the individual's parameter vector, and set the parameter range for the quality factor in the individual's parameter vector; when the individual's notch frequency is within the system frequency range and the quality factor is within the parameter range, the current individual is judged to meet the quality screening; when the notch frequency exceeds the system frequency range or the quality factor exceeds the parameter range, the current individual is judged to not meet the quality screening.

5. The radio frequency control method based on tuning response and harmonic optimization according to claim 1, characterized in that, Add redundant bits to the array configuration word and set a synchronous clock to latch the configuration word; assign an adjustable phase synchronous clock to the different weight capacitors in the multi-capacitor switching array, perform CRC check before writing data to the latch unit, and synchronously regenerate erroneous data.

6. The radio frequency control method based on tuning response and harmonic optimization according to claim 1, characterized in that, The parameter verification process includes: Based on the historical execution records of the target RF system, historical spurious energy parameters generated each time an RF signal is generated are collected, and an energy threshold is set for all the historical spurious energy parameters. The corrected spurious energy parameters of the corrected RF signal are extracted. When the corrected spurious energy parameters are lower than the energy threshold, the corrected RF signal is determined to be compliant with verification. When the corrected spurious energy parameters are higher than the energy threshold, the corrected RF signal is determined to be non-compliant with verification.

7. The radio frequency control method based on tuning response and harmonic optimization according to claim 6, characterized in that, The accuracy of the historical stray energy parameter samples is optimized, and the process includes: Step M1: Preprocess the collected historical spurious energy parameters, including removing invalid values, processing outliers and filling missing values, normalizing them based on the same scale, and representing the single execution process of the target radio frequency system in periodic form; Step M2: Set the historical stray energy parameters of the most recent few periods as target sample points, and set the historical stray energy parameters of the remaining periods as regular sample points. Calculate the data distance value from each target sample point to each regular sample point. Step M3: Sort all data distance values ​​according to size, set a percentage threshold for data distance values ​​from low to high, and collect and label the data distance values ​​within the percentage threshold as qualified distance values ​​after rounding to the nearest integer. Step M4: Compare the data correlation between each of the selected regular sample points with each target sample point, and replace each target sample point with the regular sample point with the highest data correlation.

8. The radio frequency control method based on tuning response and harmonic optimization according to claim 7, characterized in that, Let the results of normalizing each feature individual with zero mean for the target sample points and regular sample points be the first distribution values ​​E. ij1 Second distribution value F ij2 And if the data distance value is set to D, then the formula for calculating the data distance value D is: Where i represents the feature ordinal number, X represents the target sample point, Y represents the regular sample point, the number of features in each historical stray energy parameter is set to n, j1 represents the target sample point ordinal number, and j2 represents the regular sample point ordinal number.

9. The radio frequency control method based on tuning response and harmonic optimization according to claim 8, characterized in that, The calculation process for the data distance value includes: assuming the total number of periods up to the current time is N, and setting the feature individual representation of the target sample point as X. ij Where i represents the feature ordinal number, and j represents the sample ordinal number of the current sample point within the corresponding category of sample points; let the mean and standard deviation of each feature individual within each sample point be represented by μ. i and σ i Then μ i and σ i The calculation formulas are expressed as follows: The mean μ of each individual characteristic i The calculation formula is expressed as: , The standard deviation σ of each characteristic individual i The calculation formula is expressed as: ; For each current period, all characteristic individuals within the historical stray energy parameters are normalized using zero-mean normalization. Let the total distribution value of the processed result be represented by Z. ij Its calculation formula is expressed as: , Where μ ij σ represents the mean of the i-th feature of the j-th sample within the corresponding category. ij This represents the individual standard deviation of the i-th feature in the j-th sample within the corresponding category. Let the first distribution value E ij1 Second distribution value F ij2 The calculation process all conforms to the total distribution value Z. ij The calculation formula.

10. The radio frequency control method based on tuning response and harmonic optimization according to claim 9, characterized in that, The data correlation between target sample points and regular sample points is calculated based on cosine similarity, and the content includes: Let Y represent the characteristic individuals of the regular sample points. ij Let the number of features within each historical stray energy parameter be n, and let the similarity value be C. Then the vector dot product is expressed as: The vector norm of the features within the target sample points is expressed as: The vector norm of features within a regular sample point is expressed as: Then the formula for calculating the similarity value C is: .

Images