Design method, system and device of signal processing board card and storage medium

Abstract

The invention relates to the technical field of signal processing board cards, and discloses a signal processing board card design method, system and device and a storage medium, and the method comprises the steps: obtaining sample board card parameters, calculating the board card performance, inputting the sample board card parameters and performance data into a machine learning model for training, receiving performance requirement data inputted by a user, and obtaining the performance requirement data; and calculating the performance of the board card to be processed by using the trained model, evaluating whether the performance meets requirements, and if not, updating model parameters and recalculating the performance until the requirements are met. According to the method, the searching process of the signal processing board card parameters is accelerated through the machine learning model, and the problems that in traditional design, computing resource consumption is large, parameter optimization is difficult, and a prediction result is inaccurate are solved. Compared with the prior art, the method has the advantages that the board card performance can be efficiently and accurately calculated, the design efficiency is improved, and the method is suitable for a signal processing board card scene needing rapid design and performance optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing boards, and in particular to a design method, system, device and storage medium for a signal processing board. Background Art

[0002] At present, in the design process of signal processing boards, it is usually necessary to optimize multiple key parameters, such as noise figure, dynamic range, etc. The traditional design method mainly relies on manual debugging and engineers' accumulated experience. This method not only consumes a lot of time and manpower when designing complex signal processing boards, but also makes it difficult to ensure the optimality of design parameters.

[0003] In the prior art, a common method is to design a signal processing board based on simulation. By building a physical model of the board and simulating its performance in actual operation, designers can gradually approach the ideal board performance by repeatedly adjusting the parameters of the board, such as the performance parameters of components such as amplifiers, filters, and mixers. However, this method has significant limitations. First, the simulation process usually consumes a lot of computing resources, especially in high-frequency and large-bandwidth application scenarios, the simulation time may be very long. Secondly, simulation can only provide an approximate optimization method, and a lot of parameter debugging is still required in actual design. In addition, simulation can only gradually approach the optimal solution, and it is difficult to guarantee the global optimal solution.

[0004] The existing technology has the problems that the simulation process is time-consuming, resource-intensive and has limited parameter optimization effect, resulting in low design efficiency. Summary of the invention

[0005] The present invention provides a design method, system, device and storage medium for a signal processing board to achieve more efficient and accurate board design.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a design method of a signal processing board, comprising:

[0007] Acquire a plurality of sample board parameters, wherein the sample board parameters include frequency bandwidth, amplifier parameters, mixer parameters, and filter parameters;

[0008] Calculating the board performance including the board noise coefficient and the board dynamic range according to the sample board parameters;

[0009] Inputting the sample board parameters, the board noise coefficient and the board dynamic range into a machine learning model for training to obtain a trained machine learning model;

[0010] Receiving performance requirement data input by a user, wherein the performance requirement data includes an operating frequency range, an operating bandwidth, a target noise figure, and a target dynamic range of a signal to be processed;

[0011] Receiving the parameters of the board to be processed input by the user, inputting the parameters of the board to be processed into the trained machine learning model, and obtaining the performance of the board to be processed;

[0012] Determine whether the performance of the board to be processed meets the performance requirement data. If so, output the parameters of the board to be processed and the performance of the board to be processed; if not, update the parameters of the machine learning model, and output the updated board performance according to the updated machine learning model until the performance requirements are met.

[0013] In one embodiment, the calculating the board performance including the board noise coefficient and the board dynamic range according to the sample board parameters includes: The board noise figure is calculated using the following formula: , Among them, NF is the board noise factor, which represents the total noise level introduced by the characteristics of different components during the signal processing process; this factor combines the noise contributions of different components and directly affects the signal processing performance and quality of the board; In one embodiment, the board noise coefficient is calculated based on the sample board parameters, including: The calculation method of the board noise figure is: NF=NF1'×T1+NF2×T2+NF3×T3+NF4×T4 Among them, NF is the noise factor, NF1' is the noise factor caused by frequency conversion insertion loss; NF2 is the noise factor of the low noise amplifier; NF3 is the noise factor of the mixer; NF4 is the noise factor of the filter; T1 is the influence weight of frequency conversion insertion loss; T2 is the influence weight of the low noise amplifier on the noise factor; T3 is the influence weight of the mixer on the noise factor; T4 is the influence weight of the filter on the noise factor.

[0014] In one implementation, the variable frequency insertion loss influence weight is calculated as follows: T1=B1×α+B2×β+B3×γ Among them, α represents the influence coefficient of insertion loss evaluation complexity B1 in insertion loss influence weight T1; β represents the influence coefficient of matching coefficient B2 between RF chip model and operating frequency in insertion loss influence weight T1; γ represents the influence coefficient of adaptation coefficient B3 between operating frequency and filter size in insertion loss influence weight T1; B1 is the complexity of insertion loss evaluation, B2 is the matching coefficient between RF chip model and operating frequency, and B3 is the adaptation coefficient between operating frequency and filter size, among which: B1=M1×N1×W1 Among them, B1 is the complexity of insertion loss evaluation, M1 is the bandwidth percentage at the operating frequency, N1 is the percentage of frequency conversion insertion loss at the operating frequency, and W1 is the weight coefficient of the RF chip model; B2 = n1 / n2 Among them, B2 is the matching coefficient of the RF chip model and the operating frequency, n1 and n2 are the characteristic values ​​of the RF chip model and the operating frequency respectively; B3 = m1 / m2 Among them, B3 is the adaptation coefficient of the operating frequency and the filter size, m1 is the bandwidth percentage at the operating frequency, and m2 is the percentage of the filter size at the operating frequency.

[0015] In one implementation, the weight of the mixer's influence on the noise figure is calculated as follows: T3=(C1+C2+C3+C4+C5)×V1 Among them, T3 is the influence weight of the mixer on the noise coefficient, V1 is the size of the output IP impedance of the IF chip, C1 is the influence weight of the IF chip model on the IF insertion loss, C2 is the influence weight of the operating frequency on the IF insertion loss, C3 is the influence weight of the working bandwidth on the IF insertion loss, C4 is the complexity of the IF insertion loss evaluation, and C5 is the influence weight of the output impedance on the IF insertion loss.

[0016] In one implementation, the dynamic range of the board is calculated as follows: DR=DR1-NF-NF3 Wherein, DR is the dynamic range of the board, NF is the noise factor of the board, DR1 is the dynamic range calculated by the frequency bandwidth, NF3 is the influence weight of the intermediate frequency insertion loss, and the calculation formula of DR1 is: DR1 = DR2 × W2 + DR3 × W3 Among them, DR2 is the weight coefficient of ADC quantization accuracy, DR3 is the weight coefficient of ADC sampling rate; W2 is the influence weight of DR2 in dynamic range calculation, W3 is the influence weight of DR3 in dynamic range calculation, and the calculation formulas of DR2 and DR3 are as follows: DR2 = (P1a + P2) × V2, DR3 = (P1b + P3) × V3 Among them, V2 and V3 are the influence weights of ADC quantization accuracy and working bandwidth respectively, P1a and P2 are the quantitative relationships between ADC quantization accuracy and working bandwidth respectively, P1a reflects the influence of ADC quantization accuracy on DR2 (dynamic range related to quantization accuracy), P2 is a parameter related to working bandwidth, which is used to describe the influence of working bandwidth on DR2, and it reflects the influence weight of working bandwidth in dynamic range; P1b and P3 are the quantitative relationships between ADC sampling rate and working bandwidth respectively, P1b reflects the influence of ADC sampling rate on DR3 (dynamic range related to sampling rate), P3 is a parameter related to working bandwidth, which is used to describe the influence of working bandwidth on DR3, and it reflects the influence weight of working bandwidth on dynamic range related to ADC sampling rate.

[0017] In one embodiment, inputting the sample board parameters, the board noise coefficient, and the board dynamic range into a machine learning model for training to obtain a trained machine learning model includes:

[0018] Dividing the sample board parameters, the board noise coefficient and the board dynamic range into a training set and a test set;

[0019] Normalizing the training set;

[0020] Build multiple machine learning models;

[0021] Select a machine learning model, input the normalized training set into the machine learning model, and establish a mapping relationship from multiple variables to single variables;

[0022] The model accuracy is evaluated by the root mean square error and the test set, and the steps of selecting a machine learning model and evaluating it are repeated until the accuracy of the machine learning model meets the requirements, thereby determining a trained machine learning model.

[0023] In a second aspect, the present invention provides a design system for a signal processing board, comprising:

[0024] A sample data acquisition module, used to acquire a plurality of sample board parameters, wherein the sample board parameters include frequency bandwidth, amplifier parameters, mixer parameters and filter parameters;

[0025] A performance calculation module, used for calculating the board performance according to the sample board parameters, wherein the board performance includes the board noise coefficient and the board dynamic range;

[0026] A model training module, used for inputting the sample board parameters, board noise coefficient and board dynamic range into a machine learning model for training to obtain a trained machine learning model;

[0027] A performance requirement acquisition module, used to receive performance requirement data input by a user, wherein the performance requirement data includes an operating frequency range, an operating bandwidth, a target noise coefficient, and a target dynamic range of a signal to be processed;

[0028] A board performance calculation module, used for receiving board parameters to be processed input by a user, inputting the board parameters to be processed into the trained machine learning model, and obtaining the performance of the board to be processed;

[0029] The performance evaluation module is used to determine whether the performance of the board to be processed meets the performance requirement data. If so, the board parameters and the performance of the board to be processed are output; if not, the parameters of the machine learning model are updated, and the updated board performance is output according to the updated machine learning model until the performance requirements are met.

[0030] In a third aspect, the present invention further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements any one of the above-mentioned signal processing board design methods when executing the computer program.

[0031] In a fourth aspect, the present invention further provides a computer-readable storage medium, the computer-readable storage medium comprising a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned signal processing board design methods.

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

[0033] The present invention discloses a design method for a signal processing board, comprising obtaining a plurality of sample board parameters, wherein the sample board parameters include frequency bandwidth, amplifier parameters, mixer parameters and filter parameters; calculating board performance including board noise coefficient and board dynamic range according to the sample board parameters; inputting the sample board parameters, the board noise coefficient and the board dynamic range into a machine learning model for training to obtain a trained machine learning model; receiving performance requirement data input by a user, wherein the performance requirement data includes an operating frequency range, an operating bandwidth, a target noise coefficient and a target dynamic range of a signal to be processed; receiving board parameters to be processed input by a user, inputting the board parameters to be processed into the trained machine learning model to obtain board performance to be processed; judging whether the board performance to be processed meets the performance requirement data, and if so, outputting the board parameters to be processed and the board performance to be processed; if not, updating the parameters of the machine learning model, and outputting the updated board performance according to the updated machine learning model until the performance requirements are met. The present invention accelerates the search process of signal processing board parameters by combining machine learning models and physical formulas, solving the problems of large computing resource consumption, difficult parameter optimization, and insufficient data and poor interpretability of machine learning models in traditional design methods, thereby achieving more efficient and accurate signal processing board design. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic flow chart of a design method of a signal processing board provided by a first embodiment of the present invention;

[0035] Figure 2 is a schematic diagram of the structure of a design system for a signal processing board provided by a second embodiment of the present invention;

[0036] Figure 3 This is a schematic diagram of the designed board. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Reference Figure 1 The first embodiment of the present invention provides a method for designing a signal processing board, comprising the following steps: S100, acquiring a plurality of sample board parameters, wherein the sample board parameters include frequency bandwidth, amplifier parameters, mixer parameters, and filter parameters; S200, calculating board performance including board noise coefficient and board dynamic range according to the sample board parameters; S300, inputting the sample board parameters, the board noise coefficient and the board dynamic range into a machine learning model for training to obtain a trained machine learning model; S400, receiving performance requirement data input by a user, wherein the performance requirement data includes an operating frequency range, an operating bandwidth, a target noise coefficient, and a target dynamic range of a signal to be processed; S500, receiving the parameters of the board to be processed input by the user, inputting the parameters of the board to be processed into the trained machine learning model, and obtaining the performance of the board to be processed; S600, determine whether the performance of the board to be processed meets the performance requirement data. If so, output the parameters of the board to be processed and the performance of the board to be processed; if not, update the parameters of the machine learning model, and output the updated board performance according to the updated machine learning model until the performance requirements are met.

[0039] In step S100, it is necessary to obtain sample board parameters. The sample board parameters include frequency bandwidth, amplifier parameters, mixer parameters and filter parameters. These parameters represent the key characteristics of board signal processing and are basic data for evaluating board performance. Exemplarily, the method for obtaining these parameters is usually to measure and record the configuration and technical specifications of the board hardware, such as the gain of the amplifier, the frequency range of the signal, the filtering characteristics, etc. The obtained sample board parameters will be used for subsequent performance calculations to provide basic data for model training.

[0040] In an optional implementation, the frequency bandwidth, amplifier parameters, mixer parameters and filter parameters include: RF chip model, filter type, IF chip model, ADC type, power supply voltage, local oscillator frequency, crystal oscillator frequency, ADC sampling rate, ADC quantization accuracy, input reference voltage, reference accuracy and output signal reference accuracy.

[0041] In step S200, the board performance including the board noise coefficient and the board dynamic range is calculated according to the sample board parameters. It should be noted that the board performance mainly includes two indicators: the board noise coefficient and the board dynamic range. The board noise coefficient is obtained by comprehensively considering the noise contribution of each component, involving the noise coefficient NF1 of the broadband amplifier, the noise coefficient NF2 of the low noise amplifier, the noise coefficient NF3 of the mixer, and the noise coefficient NF4 of the filter. The board dynamic range is calculated based on the weight of the frequency bandwidth, the quantization accuracy of the ADC, and the sampling rate, reflecting the ability of the board to process signals. The purpose of the calculation is to provide performance data for the machine learning model for training and performance optimization.

[0042] It should be noted that the calculation of board performance is a comprehensive evaluation of various parameters. Through machine learning model fitting and optimization, the model can accurately predict the performance of the board.

[0043] Preferably, the obtaining of multiple sample board parameters includes: The more sample board parameters are input, the more accurate the model prediction results will be.

[0044] In one implementation, the calculating according to the sample board parameters to obtain the board noise coefficient includes: The board noise figure is calculated using the following formula: , Among them, NF is the board noise figure, NF1 is the noise figure of the broadband amplifier, NF2 is the noise figure of the low noise amplifier, NF3 is the noise figure of the mixer, NF4 is the noise figure of the filter, and T is the comprehensive factor.

[0045] NF1 is the noise factor of a broadband amplifier, which is a measure of the contribution of a broadband amplifier to signal noise. The noise factor of a broadband amplifier is usually affected by its design and operating characteristics. A higher NF1 will increase the overall noise level of the board, thereby reducing signal quality. Broadband amplifiers are mainly used to amplify wide-band signals, so their noise has a greater impact on the overall performance of the board. NF2 is the noise factor of the low noise amplifier, which indicates the noise introduced by the low noise amplifier during signal processing. Low noise amplifiers are usually used in the initial stage of the signal chain to maintain signal integrity and reduce noise. Unlike broadband amplifiers, low noise amplifiers are specially designed to minimize noise introduction, thereby improving the quality of signal processing and the performance of the board. NF3 is the noise factor of the mixer, which indicates the contribution of the mixer to the signal noise. The mixer is used for frequency conversion, converting the signal from one frequency range to another. In this process, the mixer may introduce a large amount of noise. NF3 is used to evaluate the impact of this frequency conversion process on signal quality, ensuring that the board can maintain a low noise level when the signal frequency changes. NF4 is the noise factor of the filter, which measures the noise introduced by the filter during signal processing. The filter is used to select signals in a specific frequency range. During the signal filtering process, some noise may be introduced. NF4 is used to evaluate the impact of the filter on the overall noise level of the board to ensure that the signal remains of good quality after being processed by the filter. T is a comprehensive factor used to adjust and integrate the impact of different noise sources on the total noise figure. It takes into account the various influence weights in the signal processing chain, such as the noise performance, frequency range, gain size, etc. of each component under different working conditions. Therefore, T can be regarded as an overall trade-off and correction factor for the noise impact of each component.

[0046] In another implementation, the calculating according to the sample board parameters to obtain the board noise coefficient includes: The board noise figure is calculated using the following formula: NF=NF1'×T1+NF2×T2+NF3×T3+NF4×T4; Among them, NF is the noise factor, NF1' is the noise factor caused by frequency conversion insertion loss; NF2 is the noise factor of the low noise amplifier; NF3 is the noise factor of the mixer; NF4 is the noise factor of the filter; T1 is the influence weight of frequency conversion insertion loss; T2 is the influence weight of the low noise amplifier on the noise factor; T3 is the influence weight of the mixer on the noise factor; T4 is the influence weight of the filter on the noise factor.

[0047] Among them, NF is the board noise factor, NF1' is the noise factor caused by frequency conversion insertion loss. NF1' is mainly used to measure the noise caused by frequency conversion insertion loss (i.e., signal energy loss during frequency conversion). During signal processing, frequency conversion insertion loss will change the strength and characteristics of the signal, which may increase noise and reduce signal quality. Therefore, NF1' reflects the degree of this noise increase; NF2 is the noise factor of the low-noise amplifier, which is similar to the broadband amplifier, but is specifically designed for amplifiers with low noise characteristics. Low-noise amplifiers are usually used for initial signal processing to maintain signal integrity; NF3 is the noise factor of the mixer, which represents the noise introduced by the mixer during signal processing. The mixer may introduce large noise during frequency conversion, and this parameter is used to evaluate its impact; NF4 is the noise factor of the filter; T1 is the impact weight of frequency conversion insertion loss; T2 is the impact weight of the low-noise amplifier on the noise factor; T3 is the impact weight of the mixer on the noise factor; T4 is the impact weight of the filter on the noise factor.

[0048] In an optional implementation, the variable frequency insertion loss influence weight is calculated as follows: T1=B1×α+B2×β+B3×γ Among them, T1 is the weight of the impact of frequency conversion insertion loss on the overall noise coefficient; it combines multiple parameters to measure the signal loss and noise level introduced by the frequency conversion process to ensure the quality and performance of the signal during the frequency conversion process; α is the influence coefficient of the insertion loss evaluation complexity B1 on T1; α indicates the relative importance of B1 in the insertion loss weight calculation. The larger its value is, the more significant the impact of the insertion loss complexity on T1 is; α is used to adjust the contribution of B1 to the total variable frequency insertion loss weight; β is the influence coefficient of the RF chip model and operating frequency matching coefficient B2 on T1; β measures the relative importance of B2 in the insertion loss weight, ensuring that the matching of the RF chip and the operating frequency has an appropriate weight for the calculation of T1; β is used to adjust the contribution of the chip model and the operating frequency adaptation in the insertion loss; γ is the influence coefficient of the operating frequency and filter size adaptation coefficient B3 on T1; γ reflects the relative influence of B3 in the insertion loss weight, ensuring that the frequency and filter size matching have appropriate weight on the calculation of T1; γ is used to adjust the influence of frequency and filter size matching in variable frequency insertion loss; B1 is the complexity of insertion loss evaluation, B2 is the matching coefficient between RF chip model and operating frequency, and B3 is the adaptation coefficient between operating frequency and filter size, where: B1=M1×N1×W1 Among them, B1 is the complexity of insertion loss evaluation, which measures the performance of insertion loss in the frequency conversion process under different signal conditions, including the influence of frequency, bandwidth and signal characteristics. M1 is the bandwidth percentage at the operating frequency, which means the proportion of signal bandwidth to the entire signal range under a certain frequency condition. M1 reflects the relationship between frequency and bandwidth. The wider the bandwidth, the larger the frequency range of the signal, and the more significant the impact on insertion loss. N1 is the percentage of frequency conversion insertion loss at the operating frequency, which means the amount of insertion loss introduced by the frequency conversion process under this frequency condition. N1 measures the degree of influence of the frequency conversion process on signal quality. A larger N1 means that the frequency conversion process brings greater signal loss. W1 is the weight coefficient of the RF chip model, which reflects the insertion loss characteristics of a specific RF chip in signal transmission. Different models of RF chips have different degrees of signal loss, and W1 measures the impact of this characteristic difference on insertion loss. B2 = n1 / n2 Among them, B2 is the matching coefficient between the RF chip model and the operating frequency, n1 and n2 are the characteristic values ​​of the RF chip model and the operating frequency respectively, n1 reflects the performance parameters of the RF chip in the signal processing process, such as gain, frequency response, etc., and n2 represents the performance requirements of the chip in processing signals under the frequency condition; B2 calculates the matching degree between the RF chip model and the operating frequency. The higher the matching degree, the better the signal processing performance of the chip under the frequency condition; B3 = m1 / m2 Among them, B3 is the adaptation coefficient of the operating frequency and the filter size, which reflects the adaptation of the signal when passing through the filter, m1 is the bandwidth percentage at the operating frequency, indicating the effective bandwidth range of the signal under the frequency condition, and m2 is the percentage of the filter size at the operating frequency, indicating the matching degree between the bandwidth response range of the filter and the signal frequency. B3 measures the adaptability between the filter size and the operating frequency. The higher the matching degree, the smaller the loss of the signal during the filter processing and the better the signal quality.

[0049] In an optional implementation manner, the weight of the influence of the mixer on the noise figure is calculated as follows: T3=(C1+C2+C3+C4+C5)×V1 T3 is the weight of the mixer's influence on the noise figure, which measures the mixer's contribution to the overall noise during signal processing. When a mixer performs frequency conversion, it may introduce noise and signal loss. T3 comprehensively considers various performance indicators of the mixer in order to more accurately assess the mixer's influence on the overall noise performance of the board, ensuring that the mixer performance can be optimized to reduce noise when designing the board. V1 is the size of the output IP impedance of the IF chip; the output impedance reflects the signal output characteristics of the IF chip, and directly affects the signal stability and insertion loss; high impedance may lead to unstable signal transmission and introduce more noise; when the impedance is well matched, the signal loss is minimal, which is conducive to maintaining signal quality; the role of V1 is to evaluate the impedance matching degree of the IF chip to ensure that the noise introduced during signal transmission is minimized and the performance of the IF chip is optimized; C1 is the weight of the impact of the IF chip model on the IF insertion loss. Different IF chip models have different performance characteristics and introduce different insertion losses. C1 is used to measure the impact of these model differences on noise, so as to guide designers to select appropriate chip models, ensure that insertion loss is minimized, and optimize signal transmission performance. C2 is the weight of the influence of the operating frequency on the intermediate frequency insertion loss. The operating frequency directly determines the spectrum range of signal transmission. High-frequency signals may be more susceptible to insertion loss. C2 is used to quantify the degree of change of insertion loss at different operating frequencies, helping designers to reasonably select the operating frequency range when designing boards, reduce noise introduction, and ensure the noise performance of boards. C3 is the weight of the effect of working bandwidth on IF insertion loss. Working bandwidth determines the range and capacity of signal transmission. A wider bandwidth may introduce more insertion loss, thus affecting noise performance. C3 is used to measure the effect of insertion loss on noise under different bandwidth conditions, helping to balance bandwidth selection and signal transmission efficiency during design, and ensuring the optimal noise performance of the board within the effective bandwidth range. C4 is the complexity of IF insertion loss assessment. Insertion loss behaves differently under different signal conditions. For example, the frequency, bandwidth, and strength of the signal will affect the insertion loss. C4 is a parameter used to assess the complexity of insertion loss and quantifies the impact of these conditions on insertion loss. Its significance lies in ensuring the accuracy of insertion loss calculation, being able to better predict and control the quality of IF signals, and providing guarantee for the stability of signal processing. C5 is the weight of the impact of output impedance on IF insertion loss. Output impedance refers to the impedance characteristics of the signal when it is output from the IF chip. Different impedance matching may cause different degrees of signal loss and may even introduce reflection noise. C5 is used to evaluate the effect of impedance matching and guide designers to make impedance matching adjustments to minimize signal insertion loss and improve signal stability and transmission quality.

[0050] In step S300, the sample board parameters and performance data are input into the machine learning model for training to obtain the mapping relationship between the sample board parameters and the board performance. In terms of implementation, the board parameters to be processed are copied to the memory area, and relevant adjustment parameters such as intermediate frequency insertion loss, bandwidth impact weight, etc. are obtained; the values ​​in the memory area are adjusted according to these parameters to generate the board parameters to be processed for training the model. Through model training, a mapping relationship between sample board parameters and board performance can be established to provide support for subsequent performance prediction.

[0051] In one embodiment, inputting the sample board parameters, the board noise coefficient, and the board dynamic range into a machine learning model for training to obtain a trained machine learning model includes:

[0052] Divide the sample board parameters, board noise coefficient and board dynamic range into a training set and a test set;

[0053] Normalize the training set to eliminate the dimensional differences between different parameters, ensure that the influence of each feature is balanced during model training, and avoid deviations caused by differences in feature value ranges; normalization is to scale the value ranges of different parameters to a unified scale, generally to a decimal between [0, 1] or (-1, 1). Since the value ranges of different board parameters (such as frequency, gain, noise coefficient, etc.) vary greatly, normalization helps to eliminate these dimensional differences, making the model more stable during training, improving training speed, and preventing the model from deviating on features with large parameter differences.

[0054] Build multiple machine learning models for training sample data; select the model that best fits the relationship between sample board parameters and performance by building models of different types or structures; build multiple machine learning models, which can include linear regression, decision trees, random forests, support vector machines (SVM), neural networks, etc. Different models have their own advantages and disadvantages. Linear models are suitable for simple situations with obvious linear relationships, while nonlinear models such as neural networks and random forests are more suitable for situations with complex parameter relationships. By building multiple models, you can compare different models and select the model that best fits the relationship between board parameters and performance.

[0055] Select a model for training, input the normalized training set data into the machine learning model, and establish a mapping relationship from the multivariate of sample board parameters to the single variable of board performance; select a machine learning model, input the normalized training set into the machine learning model, and establish a mapping relationship from multivariate to single variable. The training process is to establish a mapping relationship from multivariate to single variable, and learn the mapping between board parameters (such as frequency, gain, noise figure, etc.) and target performance indicators (noise figure and dynamic range). The training process iteratively adjusts the parameters of the model to minimize the error between the predicted value and the true value, and build a model that can accurately predict board performance.

[0056] Evaluate the accuracy of the model, evaluate the accuracy of the model prediction through the root mean square error and the test set data, and the root mean square error reflects the average deviation between the model prediction result and the true result. Evaluate the accuracy of the model through the root mean square error and the test set, and use the root mean square error (RMSE) as an indicator to evaluate the prediction error of the model. The root mean square error reflects the average deviation between the model prediction result and the true value. The smaller the value, the higher the prediction accuracy of the model. The data in the test set does not participate in the training of the model, so it is possible to objectively evaluate the generalization ability of the model for unseen data. Repeat the steps of selecting a machine learning model and evaluating it until the accuracy of the machine learning model meets the requirements, and determine the trained machine learning model. This process is an iterative optimization process: adjust the model structure, parameters, feature processing methods, etc., until the root mean square error of a model meets the requirements, or the accuracy on the test set reaches the expected standard, and finally determine the optimal trained model.

[0057] In step S400, performance requirement data input by the user is received. The performance requirement data includes the operating frequency range, operating bandwidth, target noise figure and target dynamic range of the signal to be processed. These requirements will be used as the standard for model prediction to ensure that the performance of the board design meets the user's needs. The performance requirements are usually provided through the user input interface and used as the model prediction target.

[0058] In step S500, the parameters of the board to be processed are input into the trained model to calculate the performance of the board to be processed. Based on the previous training results, the model uses the parameters of the board to be processed to quickly calculate performance indicators, including noise coefficient and dynamic range, to meet the user's performance requirements.

[0059] Preferably, the receiving of the to-be-processed board parameter input by the user, inputting the to-be-processed board parameter into the trained machine learning model, and obtaining the to-be-processed board performance includes:

[0060] Input the parameters of the board to be processed into the trained machine learning model to obtain the output performance of the board to be processed;

[0061] The performance of the board to be processed is denormalized to obtain the performance of the board to be processed.

[0062] In step S600, it is determined whether the performance of the board to be processed meets the performance requirements of the user. If the performance meets the requirements, the process proceeds to step S800 to output the parameters and performance results of the board to be processed; if it does not meet the requirements, the process proceeds to step S700 to update the model parameters and recalculate the board performance until it meets the requirements. The purpose of this step is to ensure that the final board performance meets the user's target requirements.

[0063] It should be noted that the determination of whether the performance of the board to be processed meets the performance requirements of the user includes the following steps:

[0064] The sample board parameters, the board noise coefficient and the board dynamic range are divided into a training set and a test set. The training set is used to train the machine learning model and build a mapping relationship between the sample board parameters and the board performance; the test set is used to verify the accuracy and generalization ability of the model. The division ratio is usually 70%-80% of the data as the training set and the remaining 20%-30% as the test set. Reasonable division helps ensure that the model has sufficient data during training and can effectively evaluate the prediction performance of the model during testing.

[0065] In step S700, the board design is adjusted by updating the model parameters and recalculating the board performance to gradually meet the user's performance requirements. The model parameters are optimized according to the deviation of the calculation results until the performance meets the requirements.

[0066] In one embodiment, the updating of model parameters and recalculating board performance to adjust board design to gradually meet user performance requirements includes:

[0067] Obtaining an output result of the machine learning model, wherein the output result includes a performance prediction of a parameter of a board to be processed;

[0068] According to the output results, the second board parameters are calculated, including broadband amplifier parameters, low noise amplifier parameters, image filter mixer parameters and quadrature mixer parameters, so as to further optimize the board performance;

[0069] Based on the second board parameters, adjust the parameters of the machine learning model, update and generate a new second machine learning model, and ensure that the model can adapt to changes in board performance;

[0070] The second board parameters are input into the new second machine learning model to obtain the second board performance, and the cycle is repeated until the performance requirements are met to ensure that the predetermined performance requirements are met and the board design continues to be optimized.

[0071] In step S800, the board parameters and performance results that meet the user's performance requirements are output to complete the design process of the signal processing board. This step ensures that the user can obtain board parameters and performance indicators that meet the requirements, and ultimately apply them to the actual production and use of the board.

[0072] In order to facilitate the understanding of the present invention, some preferred embodiments of the present invention are further described below.

[0073] In another implementation, the board parameters to be processed may be directly selected from a pre-stored board parameter file and copied to a memory area to serve as basic data for model training.

[0074] The working process of the present invention is described below using a relatively common scenario as an example.

[0075] Assume that the performance requirements input by a user are an operating frequency range of 0.5 GHz to 2 GHz, a bandwidth of 1 GHz, a noise figure less than 2.5, a dynamic range requirement greater than 45 dB, and an output signal reference accuracy of 14 bits.

[0076] The sample board parameters and predicted board performance manually entered according to the technical data are shown in Table 1 below:

[0077] By manually inputting the sample board parameters obtained from the technical data, the noise coefficient and dynamic range are obtained according to the corresponding calculation formula of the present invention as shown in Tables 2 and 3 below: ,

[0078] The machine learning model training obtained by linear least squares method is shown in Table 4 below:

[0079] This form includes:

[0080] Operating frequency range, bandwidth: These are performance parameters that are input to the model and represent the user’s performance requirements;

[0081] Noise figure prediction, dynamic range prediction: The model predicts these performance indicators to meet the user's target requirements;

[0082] Output signal reference accuracy prediction: The board output accuracy predicted by the model meets user requirements;

[0083] Root mean square error: an indicator to measure the prediction error of the model, the smaller the value, the better;

[0084] Training set and test set accuracy: indicates the prediction accuracy of the model on the training set and test set;

[0085] Best model selection: the best performing model during training, such as a linear regression model;

[0086] Model training time: The time taken to train the model.

[0087] It can be seen that the results predicted by the machine learning model are close to the results of manual calculation, and the model learning prediction results are closer to the user's performance requirements. Therefore, when the user inputs new board parameters, such as an operating frequency range of 1GHz to 2GHz, a bandwidth of 1GHz, a noise figure less than 2.5, a dynamic range requirement greater than 45dB, and an output signal reference accuracy of 14bit, the output result predicted by the machine learning model is greater than 2.5dB and less than 3.5dB, and the design parameters of the signal processing board can be output.

[0088] In summary, the present invention provides a method for designing a signal processing board, by obtaining sample board parameters, calculating board performance, and inputting these parameters and performance data into a machine learning model for training to establish a mapping relationship between parameters and performance, and finally calculating the performance of the board to be processed. The method realizes rapid search and performance optimization of signal processing board parameters. The method solves the problems of high computing resource consumption, difficulty in parameter optimization, and inaccurate prediction results in traditional designs. Compared with the prior art, the present invention can calculate board performance more efficiently and accurately, improve design efficiency, and is particularly suitable for signal processing board scenarios that require rapid design and performance optimization. Through training and updating, the machine learning model continuously improves the mapping relationship between board parameters and performance, ensuring the accuracy and reliability of the design results.

[0089] Reference Figure 2 A second embodiment of the present invention provides a design system for a signal processing board, comprising:

[0090] A sample data acquisition module is used to obtain sample board parameters, including frequency bandwidth, amplifier parameters, mixer parameters, filter parameters, etc.;

[0091] The performance calculation module is used to calculate the performance of the board according to the sample board parameters to obtain performance indicators such as noise coefficient and dynamic range;

[0092] A model training module is used to input sample board parameters and board performance into a machine learning model for training, so as to establish a mapping relationship between the sample board parameters and the board performance, and obtain a trained model;

[0093] A performance requirement acquisition module is used to receive performance requirements input by a user, including an operating frequency range, an operating bandwidth, a target noise figure, and a target dynamic range of a signal to be processed;

[0094] The board performance calculation module is used to input the board parameters to be processed input by the user into the trained machine learning model, calculate the performance of the board to be processed, and obtain the corresponding noise coefficient and dynamic range;

[0095] The performance evaluation module is used to evaluate whether the performance of the board to be processed meets the user's performance requirements; if so, the board parameters and performance results to be processed are output; if not, the model parameters are updated and the performance is recalculated until the requirements are met.

[0096] Preferably, the board parameters to be processed input by the user are input into the trained machine learning model, and the board parameters to be processed can also be directly selected from a pre-stored board parameter file and copied to the memory area as basic data for model training.

[0097] Preferably, the input of the board parameters to be processed input by the user into the trained machine learning model also includes input signal processing software, so that the designer can input the performance requirements in real time, obtain the corresponding board parameters, and automatically calculate and evaluate the board performance. This process can automatically and quickly match the performance indicators required by the user. If the performance requirements are met, the corresponding performance indicators are output. If the performance does not meet the requirements, the output results of the trained machine learning model will be obtained, and new prediction results will be obtained until the performance requirements input by the user are met.

[0098] It should be noted that the signal processing board design system provided in the embodiment of the present invention is used to execute all the process steps of the signal processing board design method of the above embodiment, and the working principles and beneficial effects of the two correspond one to one, so they will not be repeated.

[0099] The embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a sample data acquisition program and a performance calculation program. When the processor executes the computer program, the steps in the above-mentioned signal processing board design method embodiments are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are implemented, such as the performance calculation module.

[0100] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0101] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or a combination of certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0102] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0103] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0104] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0105] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0106] The specific embodiments described above further illustrate the purpose, technical solutions 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. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A design method for a signal processing board, characterized in that: Executed by a computer, including: Acquire a plurality of sample board parameters, wherein the sample board parameters include frequency bandwidth, amplifier parameters, mixer parameters, and filter parameters; Calculating the board performance including the board noise coefficient and the board dynamic range according to the sample board parameters; Inputting the sample board parameters, the board noise coefficient and the board dynamic range into a machine learning model for training to obtain a trained machine learning model; Receiving performance requirement data input by a user, wherein the performance requirement data includes an operating frequency range, an operating bandwidth, a target noise figure, and a target dynamic range of a signal to be processed; Receiving the parameters of the board to be processed input by the user, inputting the parameters of the board to be processed into the trained machine learning model, and obtaining the performance of the board to be processed; Determine whether the performance of the board to be processed meets the performance requirement data. If so, output the parameters of the board to be processed and the performance of the board to be processed; if not, update the parameters of the machine learning model, and output the updated board performance according to the updated machine learning model until the performance requirements are met.

2. The design method of the signal processing board according to claim 1, characterized in that: The step of calculating the board noise coefficient according to the sample board parameters includes: The board noise figure is calculated using the following formula: ， Among them, NF is the board noise figure, NF1 is the noise figure of the broadband amplifier, NF2 is the noise figure of the low noise amplifier, NF3 is the noise figure of the mixer, NF4 is the noise figure of the filter, and T is the comprehensive factor.

3. The design method of the signal processing board according to claim 1, characterized in that: The step of calculating the board noise coefficient according to the sample board parameters includes: The board noise figure is calculated using the following formula: NF=NF1'×T1+NF2×T2+NF3×T3+NF4×T4; Among them, NF is the noise factor, NF1' is the noise factor caused by frequency conversion insertion loss; NF2 is the noise factor of the low noise amplifier; NF3 is the noise factor of the mixer; NF4 is the noise factor of the filter; T1 is the influence weight of frequency conversion insertion loss; T2 is the influence weight of the low noise amplifier on the noise factor; T3 is the influence weight of the mixer on the noise factor; T4 is the influence weight of the filter on the noise factor.

4. The design method of the signal processing board according to claim 3, characterized in that: The calculation method of the variable frequency insertion loss influence weight is: T1=B1×α+B2×β+B3×γ Among them, α represents the influence coefficient of insertion loss evaluation complexity B1 in insertion loss influence weight T1; β represents the influence coefficient of matching coefficient B2 between RF chip model and operating frequency in insertion loss influence weight T1; γ represents the influence coefficient of adaptation coefficient B3 between operating frequency and filter size in insertion loss influence weight T1; B1 is the complexity of insertion loss evaluation, B2 is the matching coefficient between RF chip model and operating frequency, and B3 is the adaptation coefficient between operating frequency and filter size, among which: B1=M1×N1×W1 Among them, B1 is the complexity of insertion loss evaluation, M1 is the bandwidth percentage at the operating frequency, N1 is the percentage of frequency conversion insertion loss at the operating frequency, and W1 is the weight coefficient of the RF chip model; B2 = n1 / n2 Among them, B2 is the matching coefficient of the RF chip model and the operating frequency, n1 and n2 are the characteristic values ​​of the RF chip model and the operating frequency respectively; B3 = m1 / m2 Among them, B3 is the adaptation coefficient between the operating frequency and the filter size, m1 is the bandwidth percentage at the operating frequency, and m2 is the percentage of the filter size at the operating frequency.

5. The design method of the signal processing board according to claim 3, characterized in that: The calculation method of the influence weight of the mixer on the noise figure is: T3=(C1+C2+C3+C4+C5)×V1 Among them, T3 is the influence weight of the mixer on the noise coefficient, V1 is the size of the output IP impedance of the IF chip, C1 is the influence weight of the IF chip model on the IF insertion loss, C2 is the influence weight of the operating frequency on the IF insertion loss, C3 is the influence weight of the working bandwidth on the IF insertion loss, C4 is the complexity of the IF insertion loss evaluation, and C5 is the influence weight of the output impedance on the IF insertion loss.

6. The method for designing a signal processing board according to claim 1, characterized in that: The method of calculating the dynamic range of the board according to the sample board parameters includes: DR=DR1-NF-NF3 Wherein, DR is the dynamic range of the board, NF is the noise factor of the board, DR1 is the dynamic range calculated by the frequency bandwidth, NF3 is the influence weight of the intermediate frequency insertion loss, and the calculation formula of DR1 is: DR1 = DR2 × W2 + DR3 × W3 Among them, DR2 is the weight coefficient of ADC quantization accuracy, DR3 is the weight coefficient of ADC sampling rate; W2 is the influence weight of DR2 in dynamic range calculation, W3 is the influence weight of DR3 in dynamic range calculation, and the calculation formulas of DR2 and DR3 are as follows: DR2 = (P1a + P2) × V2, DR3 = (P1b + P3) × V3 Among them, V2 and V3 are the influence weights of ADC quantization accuracy and working bandwidth respectively, P1a and P2 are the quantitative relationship between ADC quantization accuracy and working bandwidth respectively, and P1b and P3 are the quantitative relationship between ADC sampling rate and working bandwidth respectively.

7. The design method of the signal processing board according to claim 1, characterized in that: The step of inputting the sample board parameters, the board noise coefficient, and the board dynamic range into a machine learning model for training to obtain a trained machine learning model includes: Dividing the sample board parameters, the board noise coefficient and the board dynamic range into a training set and a test set; Normalizing the training set; Build multiple machine learning models; Select a machine learning model, input the normalized training set into the machine learning model, and establish a mapping relationship from multiple variables to single variables; The model accuracy is evaluated by the root mean square error and the test set, and the steps of selecting a machine learning model and evaluating it are repeated until the accuracy of the machine learning model meets the requirements, thereby determining a trained machine learning model.

8. A design system for a signal processing board, characterized in that: include: A sample data acquisition module, used to acquire a plurality of sample board parameters, wherein the sample board parameters include frequency bandwidth, amplifier parameters, mixer parameters and filter parameters; A performance calculation module, used for calculating the board performance according to the sample board parameters, wherein the board performance includes the board noise coefficient and the board dynamic range; A model training module, used for inputting the sample board parameters, board noise coefficient and board dynamic range into a machine learning model for training to obtain a trained machine learning model; A performance requirement acquisition module, used to receive performance requirement data input by a user, wherein the performance requirement data includes an operating frequency range, an operating bandwidth, a target noise coefficient, and a target dynamic range of a signal to be processed; A board performance calculation module, used for receiving board parameters to be processed input by a user, inputting the board parameters to be processed into the trained machine learning model, and obtaining the performance of the board to be processed; The performance evaluation module is used to determine whether the performance of the board to be processed meets the performance requirement data. If so, the board parameters and the performance of the board to be processed are output; if not, the parameters of the machine learning model are updated, and the updated board performance is output according to the updated machine learning model until the performance requirements are met.

9. An electronic device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the method for designing a signal processing board as claimed in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for designing a signal processing board as claimed in any one of claims 1 to 7.

Images