A method, apparatus, medium, and product for determining parameters of an analog circuit.

Abstract

This invention discloses a method, apparatus, medium, and product for determining parameters of analog circuits, relating to the field of integrated circuit design technology. By using local performance feedback and global topology awareness as input parameters, the targetedness and generalization ability of parameter adjustment are significantly improved. A parameter candidate set for the target analog circuit is obtained through a policy function. Multiple sampling processes reduce the variance of policy gradient estimation, improving training stability and convergence speed. No dependence on a value network is required, reducing training resource overhead. The parameter candidate set output by the policy function is applied to the simulation processing scenario, ensuring the accuracy of simulation processing evaluation. Target performance indicators are extracted from the simulation results, achieving performance indicator optimization. Multiple target performance indicators are processed according to a multi-objective reward function to achieve a balance among multiple performance indicators, obtaining the parameter combination corresponding to the highest advantage value, thus improving the reliability of the analog circuit parameter determination process.

Description

Technical Field

[0001] This invention relates to the field of integrated circuit design technology, and in particular to a method, apparatus, medium, and product for determining parameters of analog circuits. Background Technology

[0002] With the development of machine learning technology, machine learning-based analog circuit parameter optimization is replacing human experience-based optimization. In machine learning-based analog circuit parameter optimization, most methods rely on value networks. Value networks require a large amount of sample data to estimate the value of states or actions to achieve accurate convergence, leading to high training resource costs and slow convergence speed. Furthermore, by iteratively approximating the true value function within the value network, it becomes trapped in local optima, making it difficult to balance multi-objective performance metrics regarding the global optimum.

[0003] Therefore, how to save training resources, achieve fast and stable convergence speed, and improve local and global optimization are problems that urgently need to be solved by those skilled in the art. Summary of the Invention

[0004] The purpose of this invention is to provide a method, apparatus, medium, and product for determining the parameters of analog circuits, in order to solve the technical problems of high training resource consumption, slow convergence speed, and difficulty in optimizing local and global parameters.

[0005] To solve the above-mentioned technical problems, the present invention provides a method for determining the parameters of an analog circuit, comprising:

[0006] Obtain the circuit source code and performance indicators corresponding to the target analog circuit;

[0007] Based on the circuit source code, the corresponding graph embedding vector is determined, and the graph embedding vector and the performance index are input into the policy function to obtain the parameter candidate set of the target analog circuit;

[0008] Simulation processing is performed on multiple parameter combinations of the parameter candidate set to obtain corresponding simulation results, and the corresponding target performance index is extracted based on the multiple simulation results;

[0009] Multiple target performance indicators are processed according to the multi-objective reward function to obtain the corresponding advantage value, and the parameter combination corresponding to the highest advantage value is used as the final parameter of the target analog circuit.

[0010] On the one hand, the process of determining the candidate set of parameters includes:

[0011] A parameter database for the target analog circuit is pre-established; wherein the parameter database stores the parameters to be determined for the target analog circuit and their corresponding value ranges;

[0012] The range of output values ​​of the output neuron of the policy function is determined based on the parameter database.

[0013] The output value of the policy function is randomly sampled multiple times based on the graph embedding vector and the performance metric to generate the parameter candidate set.

[0014] On the other hand, simulation processing is performed on multiple parameter combinations of the parameter candidate set to obtain corresponding simulation results, including:

[0015] Multiple parameter combinations are applied sequentially to the netlist;

[0016] The simulator is invoked to perform circuit simulation on the netlist to obtain the corresponding simulation results.

[0017] On the other hand, the corresponding target performance indicators are extracted based on multiple simulation results, including:

[0018] In the performance indicators of the target analog circuit, the common performance indicators corresponding to the frequency response are determined based on the circuit's functional requirements, performance requirements, multiple parameter configurations, and sensitivity analysis.

[0019] The common performance indicators are used as the target performance indicators;

[0020] The target parameters are determined based on the target performance index from multiple simulation results.

[0021] On the other hand, multiple objective performance indicators are processed according to a multi-objective reward function to obtain corresponding advantage values, including:

[0022] Obtain the weight parameters corresponding to multiple target performance indicators;

[0023] By substituting multiple target performance indicators, weight parameters, and target parameters into the multi-target reward function, reward values ​​corresponding to multiple parameter combinations are obtained.

[0024] Determine the average and variance of the reward values ​​for multiple parameter combinations;

[0025] Using the average value and the variance value as baselines, advantage processing is performed on multiple parameter combinations to obtain the corresponding advantage value.

[0026] On the other hand, the target performance indicators are at least circuit gain, bandwidth, power consumption, and linearity. Substituting multiple target performance indicators, weight parameters, and target parameters into the multi-objective reward function yields reward values ​​corresponding to multiple parameter combinations, including:

[0027] Obtain the first weight parameter corresponding to the circuit gain index, the second weight parameter corresponding to the bandwidth index, the third weight parameter corresponding to the power consumption index, and the fourth weight parameter corresponding to the linearity index;

[0028] The first ratio value is determined based on the circuit gain index and the corresponding target parameter;

[0029] The second ratio value is determined based on the bandwidth index and the corresponding target parameters;

[0030] The third ratio value is determined based on the power consumption index and the corresponding target parameters;

[0031] The fourth ratio value is determined based on the linearity index and the corresponding target parameter;

[0032] The reward value corresponding to multiple parameter combinations is obtained by processing the first ratio value and the first weight parameter, the second ratio value and the second weight parameter, the third ratio value and the third weight parameter, and the fourth ratio value and the fourth weight parameter.

[0033] On the other hand, processing the first ratio value and the first weight parameter, the second ratio value and the second weight parameter, the third ratio value and the third weight parameter, and the fourth ratio value and the fourth weight parameter yields reward values ​​corresponding to multiple parameter combinations, including:

[0034] The first ratio value and the first weight parameter are processed to obtain the first reward value;

[0035] The second ratio value and the second weight parameter are processed to obtain the second reward value;

[0036] The third ratio value and the third weight parameter are processed to obtain the third reward value;

[0037] The fourth ratio value and the fourth weight parameter are processed to obtain the fourth reward value;

[0038] The first reward value and the second reward value are added together to obtain the fifth reward value;

[0039] The final reward value is obtained by subtracting the third and fourth reward values ​​from the fifth reward value in sequence.

[0040] On the other hand, the sum of the first weight parameter and the second weight parameter is 1; the sum of the third weight parameter and the fourth weight parameter is 1.

[0041] On the other hand, the training process of the policy function includes:

[0042] The initial policy function is invoked, and the current performance metric and the graph embedding vector are input into the initial policy function; wherein, the current performance metric is the target performance metric obtained during the previous training of the policy function, and is the initial performance metric during the first training.

[0043] Output the current parameter candidate set corresponding to the initial strategy function;

[0044] The current parameter candidate set is combined with multiple parameters to perform simulation processing to obtain the corresponding current simulation result, so as to extract the corresponding current target performance index;

[0045] The performance indicators of multiple current objectives are processed according to the multi-objective reward function to obtain the corresponding current advantage value;

[0046] The initial policy function is fine-tuned based on the current advantage value to obtain the policy function after parameter fine-tuning.

[0047] When the current iteration count reaches the preset iteration count, the initial strategy function is used as the final strategy function.

[0048] If the current iteration count has not reached the preset iteration count, the strategy function after parameter fine-tuning is used as the new strategy function, and the process returns to the step of inputting the current performance index and the graph embedding vector into the initial strategy function, until the current iteration count reaches the preset iteration count, and the final strategy function is determined.

[0049] On the other hand, using the mean and variance as baselines, advantage processing is applied to multiple parameter combinations to obtain corresponding advantage values, including:

[0050] Subtract the current reward value from the average value to obtain the sixth reward value;

[0051] The square root of the variance is divided by the sixth reward value to obtain the corresponding advantage value.

[0052] On the other hand, determining the corresponding graph embedding vector based on the circuit source code includes:

[0053] Convert the circuit source code into a netlist;

[0054] The circuit diagram of the target analog circuit is constructed based on the netlist;

[0055] The circuit diagram is input into the graph neural network model to obtain the corresponding graph embedding vector.

[0056] On the other hand, the circuit diagram of the target analog circuit is constructed based on the netlist, including:

[0057] Extract the netlist nodes and netlist edge information corresponding to the netlist, wherein the netlist nodes represent the components of the target analog circuit, and the netlist edge information represents the connection relationship between the components;

[0058] The circuit diagram is obtained by processing the netlist nodes and netlist edge information using a graph data structure.

[0059] To solve the above-mentioned technical problems, the present invention also provides a parameter determination device for analog circuits, comprising:

[0060] Memory, used to store computer programs;

[0061] A processor, used to implement the steps of the parameter determination method for the analog circuit as described, when executing the computer program.

[0062] To address the aforementioned technical problems, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the parameter determination method for the analog circuit as described above.

[0063] To address the aforementioned technical problems, the present invention also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the parameter determination method for the analog circuit.

[0064] The beneficial effects of this invention are as follows: First, by determining the corresponding graph embedding vector based on the circuit source code, global topology awareness is achieved, where the graph vector corresponding to the circuit topology graph is embedded. Inputting the graph embedding vector and performance indicators into the policy function, and using local performance feedback and global topology awareness as input parameters, significantly improves the targeting and generalization ability of parameter adjustment. Second, through the policy function, a parameter candidate set for the target analog circuit is obtained. Multiple sampling processes reduce the variance of policy gradient estimation, improving training stability and convergence speed. Simultaneously, it eliminates the need for a value network, reducing training resource overhead. Third, simulation processing is performed on multiple parameter combinations from the parameter candidate set to obtain corresponding simulation results. The parameter candidate set output by the policy function is applied to the simulation processing scenario, ensuring the accuracy of simulation processing evaluation. Fourth, the target performance indicators are extracted from the simulation results, achieving performance indicator optimization. Multiple target performance indicators are processed according to a multi-objective reward function to achieve a balance among multiple performance indicators, obtaining the parameter combination corresponding to the highest advantage value. This improves both local and global optimization while also enhancing the reliability of the analog circuit parameter determination process.

[0065] Secondly, the process of determining the graph embedding vector based on the circuit source code facilitates subsequent circuit analysis, simulation of circuit parameter optimization scenarios, and data processing. The generation process of the parameter candidate set enables policy optimization through intra-group comparisons, reducing the variance of policy updates. Based on the relative performance of multiple samples rather than the absolute performance of a single sample, policy updates are more stable. This facilitates subsequent parallel processing, reduces dependence on independent value functions, and thus reduces computational resource consumption. The simulation process, through parallel simulation, eliminates the need for manual parameter adjustments, significantly shortening design verification time. It allows for rapid verification of circuit performance under different parameter configurations, quick identification and adjustment of potential problems, and reduction of circuit power consumption and cost. The determination of target performance indicators using multiple analysis methods among multiple performance metrics improves the fairness and authority of performance indicator selection, facilitating subsequent processing of multi-objective reward functions. The construction of the multi-objective reward function incorporates an adaptive weight adjustment mechanism, balancing trade-offs among multiple target performance indicators. The determination of the advantage value reduces the variance of policy updates; the baseline adjusts the reward value, making its deviation from the baseline (i.e., the advantage value) more stable. This adjustment reduces update bias caused by random fluctuations, thereby improving optimization efficiency. It allows policy updates to focus more on parameter combinations that significantly outperform the baseline, thus accelerating the algorithm's convergence. By prioritizing updates to better-performing parameter combinations, the algorithm can find the optimal solution more quickly.

[0066] In addition, the present invention also provides a parameter determination device for analog circuits, a computer-readable storage medium, and a computer program product, which have the same beneficial effects as the parameter determination method for analog circuits described above. Attached Figure Description

[0067] To more clearly illustrate the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 A flowchart illustrating a method for determining parameters of an analog circuit, provided as an embodiment of the present invention;

[0069] Figure 2 A flowchart illustrating the training process of a policy function provided in an embodiment of the present invention;

[0070] Figure 3 A structural diagram of a parameter determination device for an analog circuit provided in an embodiment of the present invention;

[0071] Figure 4This is a structural diagram of a parameter determination device for an analog circuit provided in an embodiment of the present invention. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of the present invention.

[0073] The core of this invention is to provide a method, apparatus, medium, and product for determining the parameters of analog circuits, in order to solve the technical problems of high training resource consumption, slow convergence speed, and difficulty in optimizing local and global aspects.

[0074] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0075] Traditional analog integrated circuit parameter design relies on manual selection and debugging based on the designer's experience, a time-consuming, labor-intensive process highly dependent on design experience. With the development of machine learning technology, machine learning-based analog circuit parameter optimization methods have gained widespread attention in recent years. Existing research typically transforms the circuit parameter optimization problem into a black-box optimization problem, employing methods such as genetic algorithms, Bayesian optimization, particle swarm optimization, simulated annealing, and deep learning for solution. For example, the analog circuit is treated as a black-box function of input parameters to the simulator, and parameters are adjusted based on the simulation output to improve circuit performance. Traditional reinforcement learning methods have also been used for analog circuit parameter optimization. However, traditional reinforcement learning algorithms rely on value networks in the calculation of the advantage function, which not only increases training resource overhead but may also lead to estimation errors; their sample efficiency is low in high-dimensional cases, and their convergence speed is slow. The analog circuit parameter determination method provided by this invention can solve the above-mentioned technical problems.

[0076] Figure 1 A flowchart of a method for determining parameters of an analog circuit provided in an embodiment of the present invention is shown below. Figure 1 As shown, the method includes:

[0077] S11: Obtain the circuit source code and performance indicators corresponding to the target analog circuit;

[0078] S12: Determine the corresponding graph embedding vector based on the circuit source code, and input the graph embedding vector and performance indicators into the strategy function to obtain the parameter candidate set of the target analog circuit;

[0079] S13: Perform simulation processing on multiple parameter combinations of the parameter candidate set to obtain the corresponding simulation results, and extract the corresponding target performance index based on the multiple simulation results;

[0080] S14: Process multiple target performance indicators according to the multi-objective reward function to obtain the corresponding advantage value, and take the parameter combination corresponding to the highest advantage value as the final parameters of the target analog circuit.

[0081] Specifically, the circuit source code in step S11 is the code that describes the behavior of digital circuits and systems, and the source code that implements the circuit simulation function. It can be obtained through an open-source digital circuit simulator or by using other programming languages, and there is no limitation here.

[0082] Performance metrics are quantitative indicators used to measure the performance of a specific system, project, or activity under a target analog circuit. They help assess whether the system or process has achieved the expected performance level. These mainly include metrics such as DC operating point, AC frequency response, transient response, noise analysis, and power consumption. There can be one or more performance metrics; to achieve a balance between these metrics, this application employs multiple performance metrics.

[0083] In step S12, the graph embedding vector is determined based on the circuit source code. The circuit source code is converted into a netlist using open-source tools, the circuit diagram is constructed using the netlist, and then the circuit diagram is mapped to a low-dimensional embedding vector. This avoids conversion when performing policy functions, and ensures that the input format of the policy functions is matched.

[0084] The graph embedding vector and performance metrics are input into the policy function, enabling the policy function to both perceive the physical connections of the circuit and optimize in real time based on performance feedback. The resulting parameter candidate set for the target analog circuit is characterized by a reinforcement learning algorithm that does not rely on a value model. This can be an optimization policy function that directly optimizes behavior selection by adjusting the policy; it can also optimize the policy by learning action value functions; or it can optimize the policy by comparing the reward values ​​of multiple actions in a unified state, without relying on an independent value network. Regarding the policy function, a Group Relative Policy Optimization (GRPO) function can be used, sampling multiple times within each group of parameters to be optimized to obtain the parameter candidate set. It should be noted that the parameter candidate set contains G different circuit parameter configurations.

[0085] Step S13 involves simulating multiple parameter combinations from the parameter candidate set to obtain corresponding simulation results. This simulation processing primarily focuses on analog circuit simulation and can utilize analog circuit simulation software or other programming programs; no limitation is made here. The goal is to obtain the corresponding simulation results. It should be noted that, since there are multiple parameter combinations, the number of simulation results is the same as the number of parameter combinations. A parameter combination refers to the parameter configuration combination of multiple components corresponding to the target analog circuit.

[0086] Based on multiple simulation results, corresponding target performance indicators are extracted. Here, key performance indicators are identified from multiple performance indicators of the target analog circuit and used as target performance indicators. The target performance indicators here correspond to the specific parameters of the indicators.

[0087] The multi-objective reward function in step S14 corresponds to reward functions under multiple performance indicators. For each parameter configuration under these performance indicators, a reward function is set, and through processing, reward values ​​corresponding to different parameter configurations are obtained. Combined with the aforementioned G groups of different circuit parameter configurations, this corresponds to G reward values. The corresponding advantage values ​​are determined based on these reward values, resulting in G advantage values. By setting these G advantage values, the parameter combination corresponding to the highest advantage value is used as the final parameters of the target analog circuit.

[0088] The beneficial effects of this invention are as follows: First, by determining the corresponding graph embedding vector based on the circuit source code, global topology awareness is achieved, where the graph vector corresponding to the circuit topology graph is embedded. Inputting the graph embedding vector and performance indicators into the policy function, and using local performance feedback and global topology awareness as input parameters, significantly improves the targeting and generalization ability of parameter adjustment. Second, through the policy function, a parameter candidate set for the target analog circuit is obtained. Multiple sampling processes reduce the variance of policy gradient estimation, improving training stability and convergence speed. Simultaneously, it eliminates the need for a value network, reducing training resource overhead. Third, simulation processing is performed on multiple parameter combinations from the parameter candidate set to obtain corresponding simulation results. The parameter candidate set output by the policy function is applied to the simulation processing scenario, ensuring the accuracy of simulation processing evaluation. Fourth, the target performance indicators are extracted from the simulation results, achieving performance indicator optimization. Multiple target performance indicators are processed according to a multi-objective reward function to achieve a balance among multiple performance indicators, obtaining the parameter combination corresponding to the highest advantage value. This improves both local and global optimization while also enhancing the reliability of the analog circuit parameter determination process.

[0089] In some embodiments, determining the corresponding graph embedding vector based on the circuit source code includes:

[0090] Convert the circuit source code into a netlist;

[0091] The circuit diagram of the target analog circuit is constructed based on the netlist;

[0092] The graph neural network model is invoked, and the circuit diagram is input into the graph neural network model to obtain the corresponding graph embedding vector.

[0093] Specifically, the circuit source code is converted into a netlist. Circuit source code typically exists in the form of a hardware description language. This conversion is performed using logic synthesis tools or open-source tools. The netlist is then converted into a visual circuit diagram using conversion tools; the purpose of this conversion is to make the circuit diagram structure manageable. Based on the netlist, a circuit diagram of the target analog circuit is constructed, facilitating subsequent analysis and optimization.

[0094] The circuit diagram is input into the graph neural network model to obtain the corresponding graph embedding vector. The specific method for generating the graph embedding vector can be the same as or different from the conventional model generation method, and can be set according to the actual situation.

[0095] The process of determining the graph embedding vector based on the circuit source code provided in this embodiment facilitates subsequent circuit analysis, simulation of circuit parameter optimization scenarios, and data processing.

[0096] In some embodiments, the circuit diagram of the target analog circuit is constructed based on the netlist, including:

[0097] Extract the netlist nodes and netlist edge information corresponding to the netlist. The netlist nodes represent the components of the target analog circuit, and the netlist edge information represents the connection relationship between the components.

[0098] The circuit diagram is obtained by processing the information of netlist nodes and netlist edges using a graph data structure.

[0099] Specifically, a netlist is a text file that describes the connections between circuits. It mainly consists of netlist nodes, netlist edge information, and ports. Netlist nodes are components in the circuit, such as resistors, capacitors, and transistors. Netlist edge information describes the connections between components. Ports are input and output ports. The netlist file represents the circuit connections in a specific format. Component declarations include the type, name, and connection node of each component.

[0100] The netlist node and netlist edge information extracted from the netlist are used to represent the circuit using a graph data structure. The graph can be represented using an adjacency list or an adjacency matrix.

[0101] The netlist construction circuit diagram provided in this embodiment allows for a visual view of the circuit topology, the connection relationships and layout of multiple components, laying the foundation for subsequent generation of graph embedding vectors.

[0102] In some embodiments, the process of determining the parameter candidate set includes:

[0103] A parameter database for the target analog circuit is pre-established; the parameter database stores the parameters to be determined for the target analog circuit and their corresponding value ranges.

[0104] Determine the range of output values ​​for the output neurons of the policy function based on the parameter database;

[0105] The output value of the policy function is randomly sampled multiple times based on the graph embedding vector and performance metrics to generate a set of parameter candidates.

[0106] Specifically, the parameter database of the target analog circuit stores the parameters to be determined for the target analog circuit. The parameters to be determined only include their types and corresponding value ranges. Parameter configuration combinations can be determined through mapping relationships using a combination set approach or a list approach, depending on the actual situation.

[0107] The parameter database is used as a function parameter of the policy function. Here, the function parameter corresponds to the action space of the policy function, which is a constraint that the specific value of the parameter must not exceed the range of that value. The embedding vector and performance index are randomly sampled multiple times to generate a parameter candidate set, i.e., G different circuit parameter configurations.

[0108] The parameter candidate set generation process provided in this embodiment facilitates policy optimization through intra-group comparisons, reducing the variance of policy updates. By basing policy updates on the relative performance of multiple samples rather than the absolute performance of a single sample, the process becomes more stable. This also facilitates subsequent parallel processing, reduces dependence on independent value functions, and consequently lowers computational resource consumption.

[0109] In some embodiments, simulation processing is performed on multiple parameter combinations of the parameter candidate set to obtain corresponding simulation results, including:

[0110] Multiple parameter combinations are applied sequentially to the netlist;

[0111] Call the simulator to perform circuit simulation on the netlist to obtain the corresponding simulation results.

[0112] Specifically, simulation processing is performed on multiple parameter combinations, where the number of simulation results is the same as the number of parameter combinations. For each parameter combination, it is sequentially applied to the netlist, and the simulator is invoked to perform circuit simulation processing to obtain the corresponding simulation results.

[0113] The simulator here can be an integrated circuit simulation program (SPICE) simulator or other simulators, without limitation. However, in order to obtain simulation results that follow the same simulation mechanism, the same simulator is used during the simulation process.

[0114] The simulation process provided in this embodiment uses parallel simulation processing, which eliminates the need for manual parameter adjustment, greatly shortening the design verification time. It can quickly verify the circuit performance under different parameter configurations, quickly identify potential problems and make adjustments, and reduce the power consumption and cost of the circuit.

[0115] In some embodiments, the corresponding target performance index is extracted based on multiple simulation results, including:

[0116] Based on the functional requirements, performance requirements, multiple parameter configurations, and sensitivity analysis of the target analog circuit, the common performance indicators corresponding to the frequency response are determined.

[0117] Use common performance indicators as target performance indicators;

[0118] The target parameters are determined based on the target performance indicators from multiple simulation results.

[0119] Specifically, analog circuits involve many performance indicators. The key is to find the key performance indicators among them, and to select those that can comprehensively reflect the performance of the circuit and are closely related to the design goals.

[0120] Specifically, the target performance index can be determined through any one of the following analysis methods: functional requirements analysis, performance requirements analysis, combination of multiple parameter configurations, and sensitivity analysis. Alternatively, a combination of these analysis methods can be used to determine a common performance index, which can then be used as the target performance index.

[0121] Functional requirements analysis primarily focuses on components such as amplifiers, filters, and oscillators. For amplifiers, performance metrics include gain, bandwidth, input impedance, and output impedance; for filters, cutoff frequency, passband gain, and stopband attenuation; and for oscillators, oscillation frequency, phase noise, and harmonic distortion. Performance requirements analysis can include low noise requirements (noise power, signal-to-noise ratio), high bandwidth requirements (bandwidth, gain flatness), and low power consumption requirements (static power consumption, dynamic power consumption). This involves combinations of multiple parameter configurations, different component values ​​(gain and bandwidth under different resistance and capacitance values), different power supply voltages (power consumption and performance under different power supply voltages), and different temperature ranges (noise and stability under different temperatures). Sensitivity analysis identifies which performance metrics are most sensitive to parameter changes, such as the sensitivity of gain to resistance values ​​(analyzing gain changes with resistance values ​​through simulation) and the sensitivity of bandwidth to capacitance values ​​(analyzing bandwidth changes with capacitance values ​​through simulation).

[0122] Common performance indicators are those that must be considered under all the above-mentioned analysis methods, such as circuit gain, bandwidth, power consumption, and linearity. Linearity, in particular, is an important indicator that measures the degree of linearity between the output and input of a sensor or measurement system. Based on these indicators, the corresponding target parameter values, i.e., reference values, are determined.

[0123] This embodiment provides a method for determining target performance indicators based on multiple performance metrics using various analytical approaches, thereby improving the fairness and authority of performance indicator selection and facilitating subsequent processing of multi-objective reward functions.

[0124] In some embodiments, multiple objective performance metrics are processed according to a multi-objective reward function to obtain corresponding advantage values, including:

[0125] Obtain the weight parameters corresponding to multiple target performance indicators;

[0126] By substituting multiple target performance indicators, weight parameters, and target parameters into the multi-target reward function, the reward value corresponding to multiple parameter combinations is obtained.

[0127] Determine the average and variance of the reward values ​​for multiple parameter combinations;

[0128] Using the mean and variance as baselines, we perform dominance processing on multiple parameter combinations to obtain the corresponding dominance values.

[0129] Specifically, the weight parameters corresponding to the target performance indicators are obtained, and the target performance indicators, corresponding weight parameters, and target parameters are substituted into the multi-target reward function to obtain the reward value corresponding to the combination of multiple parameters.

[0130] The average and variance of the reward values ​​of multiple parameter combinations are used as a baseline, and advantage processing is performed on each parameter combination to obtain the corresponding advantage value.

[0131] The process provided in this embodiment, which processes multiple target performance indicators based on a multi-objective reward function to obtain corresponding advantage values, does not require an additional value network, thus reducing training overhead costs.

[0132] In some embodiments, the target performance metrics are at least circuit gain, bandwidth, power consumption, and linearity. Substituting multiple target performance metrics, weight parameters, and target parameters into a multi-target reward function yields reward values ​​corresponding to multiple parameter combinations, including:

[0133] Obtain the first weighted parameter corresponding to the circuit gain index, the second weighted parameter corresponding to the bandwidth index, the third weighted parameter corresponding to the power consumption index, and the fourth weighted parameter corresponding to the linearity index;

[0134] The first proportional value is determined based on the circuit gain index and the corresponding target parameter;

[0135] The second ratio value is determined based on the bandwidth index and the corresponding target parameters;

[0136] The third ratio value is determined based on the power consumption index and the corresponding target parameters;

[0137] The fourth ratio value is determined based on the linearity index and the corresponding target parameter;

[0138] The reward value corresponding to multiple parameter combinations is obtained by processing the first ratio value and the first weight parameter, the second ratio value and the second weight parameter, the third ratio value and the third weight parameter, and the fourth ratio value and the fourth weight parameter.

[0139] Specifically, weights are assigned to each of the four target performance indicators, and proportional values ​​are determined for each target performance indicator and its corresponding specific parameters. The reward value is then obtained by processing based on these proportional values ​​and weight parameters.

[0140] The process of constructing a multi-objective reward function provided in this embodiment introduces an adaptive weight adjustment mechanism to balance and trade off among multiple objective performance indicators.

[0141] In some embodiments, processing the first ratio value and the first weight parameter, the second ratio value and the second weight parameter, the third ratio value and the third weight parameter, and the fourth ratio value and the fourth weight parameter yields a reward value corresponding to multiple parameter combinations, including:

[0142] The first reward value is obtained by processing the first proportion value and the first weight parameter;

[0143] The second reward value is obtained by processing the second proportion value and the second weight parameter;

[0144] The third reward value is obtained by processing the third proportion value and the third weight parameter;

[0145] The fourth reward value is obtained by processing the fourth proportion value and the fourth weight parameter;

[0146] The first reward value and the second reward value are added together to obtain the fifth reward value;

[0147] The final reward value is obtained by subtracting the third and fourth reward values ​​from the fifth reward value in sequence.

[0148] Specifically, the formula is as follows:

[0149] ;

[0150] in, , , , These are circuit gain, bandwidth, power consumption, and linearity metrics, respectively. , , , These are the target parameters corresponding to circuit gain, bandwidth, power consumption, and linearity, respectively. , , , These correspond to the first, second, third, and fourth proportional values, respectively. , , and These correspond to the first weight parameter, the second weight parameter, the third weight parameter, and the fourth weight parameter, respectively.

[0151] The multi-objective reward function construction process provided in this embodiment achieves a flexible balance among target performance indicators such as gain, bandwidth, power consumption, and linearity by weighting different target performance indicators.

[0152] In some embodiments, the first weight parameter and the second weight parameter are summed to 1; the third weight parameter and the fourth weight parameter are summed to 1.

[0153] In the above embodiments, after the weighted sum of the first and second weight parameters, the weighted sum of the third and fourth weight parameters needs to be subtracted. This is because the reward value is combined with the consumption of power consumption and linearity indicators in the actual hardware circuit. At the same time, multiple performance indicators are optimized. Subtracting power consumption and linearity ensures that a balance is found between performance and resources during the optimization process, avoiding over-optimization of one indicator and neglecting other important indicators.

[0154] In some embodiments, the training process of the policy function includes:

[0155] Call the initial policy function, inputting the current performance metric and the graph embedding vector into the initial policy function; where the current performance metric is the target performance metric obtained during the previous training of the policy function, and is the initial performance metric during the first training.

[0156] Output the current parameter candidate set corresponding to the initial policy function;

[0157] The simulation process is performed on multiple parameter combinations in the current parameter candidate set to obtain the corresponding current simulation results, so as to extract the corresponding current target performance index.

[0158] The performance indicators of multiple current objectives are processed according to the multi-objective reward function to obtain the corresponding current advantage value;

[0159] The initial policy function is fine-tuned based on the current advantage value to obtain the policy function after parameter fine-tuning.

[0160] When the current iteration count reaches the preset iteration count, the initial policy function will be used as the final policy function.

[0161] If the current iteration count has not reached the preset iteration count, the strategy function after parameter fine-tuning is used as the new strategy function, and the process returns to the step of inputting the current performance index and the graph embedding vector into the initial strategy function, until the current iteration count reaches the preset iteration count, and the final strategy function is determined.

[0162] Specifically, in a reinforcement learning environment, a state space, action space, and reward function are defined. The state can consist of circuit performance metrics (such as current gain, bandwidth, etc.), and the action corresponds to the value of the parameter to be determined. The input layer of the policy network receives the embedding vector corresponding to the circuit plus the performance metric; the hidden layer uses a multi-layer fully connected or structured attention mechanism; the output layer predicts the circuit parameters and adjusts actions (such as bias voltage). The initial state can be randomly selected; for example, the policy network weights can be randomly initialized, and the bias term can be set to zero.

[0163] The circuit source code is converted into a netlist using open-source tools. Netlist nodes are used as graph nodes, and connections are used as graph edges to construct the circuit diagram. Then, a graph neural network is used to map the circuit diagram to a low-dimensional embedding vector.

[0164] The graph embedding vector corresponding to the circuit diagram, plus the performance metrics obtained in the previous round, are used as input. A set of candidate parameters (i.e., A different circuit parameter configurations) is randomly sampled based on the current initial policy function. The parameter values ​​are applied to the circuit netlist, and the SPICE simulator is called to perform circuit simulation to obtain the corresponding simulation results. Key performance metrics of the circuit are extracted from the simulation results, and a reward value for each parameter configuration is calculated using a pre-designed multi-objective reward function. Based on the mean and variance of the reward values ​​as a baseline, the advantage value of each parameter configuration is calculated. The initial policy function is fine-tuned according to the advantage value to determine a new policy function. The current iteration number is recorded. When the current iteration number reaches the preset iteration number, the initial policy function has been trained successfully. If the current iteration number has not reached the preset iteration number, the policy function with fine-tuned parameters is used as the new policy function, and training continues until the current iteration number reaches the preset iteration number. Training is then complete, and the final policy function is determined.

[0165] In addition to the constraint of the number of iterations, performance indicators can also be determined to meet preset requirements.

[0166] The training process of the policy function provided in this embodiment can automatically parse the circuit netlist, construct the circuit topology graph, and extract graph embeddings through a graph neural network. Simultaneously, performance metrics such as gain, bandwidth, power consumption, and linearity obtained from the previous round of SPICE simulation are used as auxiliary inputs, enabling the policy network to both perceive the physical connections of the circuit and optimize and adjust in real time based on performance feedback. After feeding multimodal inputs into the policy function, the network performs multiple intra-group samplings for each action and calculates the average reward baseline, achieving robust search in the high-dimensional parameter space. This method significantly reduces the variance of policy gradient estimation, improves training stability and convergence speed, and is more suitable for global optimization of complex analog circuits compared to other algorithms. Constructing a composite reward function and introducing an adaptive weight adjustment mechanism allows for flexible balancing of objectives such as gain, bandwidth, power consumption, and linearity.

[0167] In some embodiments, the initial policy function is fine-tuned based on the advantage value to obtain a fine-tuned policy function, including:

[0168] Substitute the advantage value into the loss function of the initial policy function, and perform gradient processing on the loss function to obtain the adjustment amount of the policy function weights and biases.

[0169] The weight and bias adjustments are applied to the initial policy function to complete the parameter fine-tuning of the initial policy function.

[0170] Specifically, by optimizing the policy function, calculating the gradient and updating the current policy function, gradient processing is performed based on the advantage value. The specific processing method can be the same as or different from the conventional gradient descent method, and can be set according to the actual situation.

[0171] The optimization strategy function provided in this embodiment can accelerate the optimization process of parameter weights and bias terms, effectively avoid getting trapped in local optima, adjust the learning rate and number of iterations to ensure that the optimal solution is searched globally, and improve the optimization quality.

[0172] In some embodiments, the mean and variance values ​​are used as baselines, and dominance processing is performed on multiple parameter combinations to obtain the corresponding dominance value, including:

[0173] Subtract the current reward value from the average value to obtain the sixth reward value;

[0174] The corresponding advantage value is obtained by dividing the square root of the variance by the sixth reward value.

[0175] The specific formula for calculating the advantage is as follows:

[0176] ;

[0177] in, This is the current reward value. This is the average value. The sixth reward value, It is the square root of the variance. Indicates the first A combination of parameters.

[0178] The advantage value determination process provided in this embodiment reduces the variance of policy updates. The baseline's role is to adjust the reward value, making its deviation from the baseline (i.e., the advantage value) more stable. This adjustment reduces update bias caused by random fluctuations, thereby improving optimization efficiency. It allows policy updates to focus more on parameter combinations that are significantly better than the baseline, thus accelerating the algorithm's convergence speed. By prioritizing updates to those better-performing parameter combinations, the algorithm can find the optimal solution faster.

[0179] Figure 2 A flowchart illustrating the training process of a policy function provided in an embodiment of the present invention is shown below. Figure 2 Shown, including:

[0180] S21: Determine the parameters to be determined and their value ranges based on the circuit structure and design requirements;

[0181] S22: Convert the circuit source code into a netlist, construct the circuit diagram, and then use a graph neural network to obtain the graph embedding vector;

[0182] S23: Initialize the weights and biases of the policy function;

[0183] S24: Take the graph embedding vector and the performance index obtained in the previous round as input, so that the policy function generates a set of parameter candidates through random sampling;

[0184] S25: Apply the parameter candidate set to the circuit netlist respectively, and call the simulator to obtain the simulation results;

[0185] S26: Extract performance indicators from simulation results and calculate reward values ​​and corresponding advantage values;

[0186] S27: Based on the loss function and advantage value, use gradient calculation to update the weights and biases of the policy function;

[0187] S30: Determine whether the number of training rounds has reached the preset threshold. If yes, end the process; otherwise, return to step S24.

[0188] The foregoing has described in detail various embodiments of the method for determining the parameters of analog circuits. Based on this, the present invention also discloses a device for determining the parameters of analog circuits corresponding to the above-described method. Figure 3 This is a structural diagram of a parameter determination device for an analog circuit provided in an embodiment of the present invention. Figure 3 As shown, the device for determining the parameters of the analog circuit includes:

[0189] Module 11 is used to acquire the circuit source code and performance indicators corresponding to the target analog circuit;

[0190] Input module 12 is used to determine the corresponding graph embedding vector based on the circuit source code, and input the graph embedding vector and performance indicators into the strategy function to obtain the parameter candidate set of the target analog circuit;

[0191] The extraction module 13 is used to perform simulation processing on multiple parameter combinations of the parameter candidate set to obtain the corresponding simulation results, and extract the corresponding target performance index based on the multiple simulation results;

[0192] The processing module 14 is used to process multiple target performance indicators according to the multi-objective reward function to obtain the corresponding advantage value, and use the parameter combination corresponding to the highest advantage value as the final parameters of the target analog circuit.

[0193] Since the embodiments of the device part correspond to the embodiments described above, please refer to the embodiments of the method part for the description of the device part, and will not be repeated here.

[0194] For a description of the parameter determination device for analog circuits provided by the present invention, please refer to the above method embodiments. The present invention will not be described in detail here, but it has the same beneficial effects as the above-described parameter determination method for analog circuits.

[0195] Figure 4 A structural diagram of a parameter determination device for an analog circuit provided in an embodiment of the present invention is shown below. Figure 4 As shown, the device includes:

[0196] Memory 21 is used to store computer programs;

[0197] Processor 22 is used to implement the steps of a method for determining the parameters of an analog circuit when executing a computer program.

[0198] The parameter determination device for the analog circuit provided in this embodiment may include, but is not limited to, smartphones, tablets, laptops, or desktop computers.

[0199] The processor 22 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 22 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array. The processor 22 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 22 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 22 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.

[0200] The memory 21 may include one or more computer-readable storage media, which may be non-transitory. The memory 21 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 21 is used to store at least the following computer program 211, which, after being loaded and executed by the processor 22, is capable of implementing the relevant steps of the analog circuit parameter determination method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 21 may also include an operating system 212 and data 213, etc., and the storage method may be temporary storage or permanent storage. The operating system 212 may include Windows, Unix, Linux, etc. The data 213 may include, but is not limited to, the data involved in the analog circuit parameter determination method, etc.

[0201] In some embodiments, the parameter determination device for analog circuits may further include a display screen 23, an input / output interface 24, a communication interface 25, a power supply 26, and a communication bus 27.

[0202] Those skilled in the field can understand, Figure 4 The structures shown do not constitute a limitation on the parameter determination device for analog circuits and may include more or fewer components than those shown.

[0203] The processor 22 implements the parameter determination method for the analog circuit provided in any of the above embodiments by calling the instructions stored in the memory 21.

[0204] For a description of the parameter determination device for analog circuits provided by the present invention, please refer to the above method embodiments. The present invention will not be described in detail here, but it has the same beneficial effects as the above-described parameter determination method for analog circuits.

[0205] Furthermore, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by processor 22, implements the steps of the parameter determination method for the analog circuit described above.

[0206] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0207] For an introduction to the computer-readable storage medium provided by the present invention, please refer to the above method embodiments. The present invention will not be described in detail here, but it has the same beneficial effects as the parameter determination method of the above analog circuit.

[0208] Furthermore, the present invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of a parameter determination method for an analog circuit.

[0209] For an introduction to the computer program product provided by the present invention, please refer to the above method embodiments. The present invention will not be described in detail here, but it has the same beneficial effects as the above-described method for determining the parameters of analog circuits.

[0210] The present invention has provided a detailed description of a method, apparatus, medium, and product for determining parameters of an analog circuit. The various embodiments in the specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make various improvements and modifications to the present invention without departing from its principles, and these improvements and modifications also fall within the protection scope of the present invention.

[0211] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

Claims

1. A method for determining parameters of an analog circuit, characterized in that, include: Obtain the circuit source code and performance indicators corresponding to the target analog circuit; Based on the circuit source code, the corresponding graph embedding vector is determined, and the graph embedding vector and the performance index are input into the policy function to obtain the parameter candidate set of the target analog circuit; Simulation processing is performed on multiple parameter combinations of the parameter candidate set to obtain corresponding simulation results, and the corresponding target performance index is extracted based on the multiple simulation results; Multiple target performance indicators are processed according to the multi-objective reward function to obtain the corresponding advantage value, and the parameter combination corresponding to the highest advantage value is used as the final parameter of the target analog circuit. Correspondingly, target performance indicators are extracted based on multiple simulation results, including: In the performance indicators of the target analog circuit, the common performance indicators corresponding to the frequency response are determined based on the circuit's functional requirements, performance requirements, multiple parameter configurations, and sensitivity analysis. The common performance indicators are used as the target performance indicators; The target parameters are determined based on the target performance index from multiple simulation results; Correspondingly, multiple objective performance indicators are processed according to the multi-objective reward function to obtain corresponding advantage values, including: Obtain the weight parameters corresponding to multiple target performance indicators; By substituting multiple target performance indicators, weight parameters, and target parameters into the multi-target reward function, reward values ​​corresponding to multiple parameter combinations are obtained. Determine the average and variance of the reward values ​​for multiple parameter combinations; Using the average value and the variance value as baselines, advantage processing is performed on multiple parameter combinations to obtain the corresponding advantage value.

2. The method for determining the parameters of an analog circuit according to claim 1, characterized in that, The process of determining the candidate set of parameters includes: A parameter database for the target analog circuit is pre-established; wherein the parameter database stores the parameters to be determined for the target analog circuit and their corresponding value ranges; The range of output values ​​of the output neuron of the policy function is determined based on the parameter database. The output value of the policy function is randomly sampled multiple times based on the graph embedding vector and the performance metric to generate the parameter candidate set.

3. The method for determining the parameters of an analog circuit according to claim 1, characterized in that, Simulation results are obtained by performing simulation processing on multiple parameter combinations of the parameter candidate set, including: Multiple parameter combinations are applied sequentially to the netlist; The simulator is invoked to perform circuit simulation on the netlist to obtain the corresponding simulation results.

4. The method for determining the parameters of an analog circuit according to claim 3, characterized in that, The target performance indicators are at least circuit gain, bandwidth, power consumption, and linearity. Substituting multiple target performance indicators, weight parameters, and target parameters into the multi-objective reward function yields reward values ​​corresponding to multiple parameter combinations, including: Obtain the first weight parameter corresponding to the circuit gain index, the second weight parameter corresponding to the bandwidth index, the third weight parameter corresponding to the power consumption index, and the fourth weight parameter corresponding to the linearity index; The first ratio value is determined based on the circuit gain index and the corresponding target parameter; The second ratio value is determined based on the bandwidth index and the corresponding target parameters; The third ratio value is determined based on the power consumption index and the corresponding target parameters; The fourth ratio value is determined based on the linearity index and the corresponding target parameter; The reward value corresponding to multiple parameter combinations is obtained by processing the first ratio value and the first weight parameter, the second ratio value and the second weight parameter, the third ratio value and the third weight parameter, and the fourth ratio value and the fourth weight parameter.

5. The method for determining the parameters of an analog circuit according to claim 4, characterized in that, The reward values ​​corresponding to multiple parameter combinations are obtained by processing the first ratio value and the first weight parameter, the second ratio value and the second weight parameter, the third ratio value and the third weight parameter, and the fourth ratio value and the fourth weight parameter, including: The first ratio value and the first weight parameter are processed to obtain the first reward value; The second ratio value and the second weight parameter are processed to obtain the second reward value; The third ratio value and the third weight parameter are processed to obtain the third reward value; The fourth ratio value and the fourth weight parameter are processed to obtain the fourth reward value; The first reward value and the second reward value are added together to obtain the fifth reward value; The final reward value is obtained by subtracting the third and fourth reward values ​​from the fifth reward value in sequence.

6. The method for determining the parameters of an analog circuit according to claim 4, characterized in that, The sum of the first weight parameter and the second weight parameter is 1; the sum of the third weight parameter and the fourth weight parameter is 1.

7. The method for determining the parameters of an analog circuit according to claim 3, characterized in that, The training process of the policy function includes: The initial policy function is invoked, and the current performance metric and the graph embedding vector are input into the initial policy function; wherein, the current performance metric is the target performance metric obtained during the previous training of the policy function, and is the initial performance metric during the first training. Output the current parameter candidate set corresponding to the initial strategy function; The current parameter candidate set is combined with multiple parameters to perform simulation processing to obtain the corresponding current simulation result, so as to extract the corresponding current target performance index; The performance indicators of multiple current objectives are processed according to the multi-objective reward function to obtain the corresponding current advantage value; The initial policy function is fine-tuned based on the current advantage value to obtain the policy function after parameter fine-tuning. When the current iteration count reaches the preset iteration count, the initial strategy function will be used as the final strategy function. If the current iteration count has not reached the preset iteration count, the strategy function after parameter fine-tuning is used as the new strategy function, and the process returns to the step of inputting the current performance index and the graph embedding vector into the initial strategy function, until the current iteration count reaches the preset iteration count, and the final strategy function is determined.

8. The method for determining the parameters of an analog circuit according to claim 3, characterized in that, Using the mean and variance as baselines, multiple parameter combinations are subjected to dominance processing to obtain corresponding dominance values, including: Subtract the current reward value from the average value to obtain the sixth reward value; The square root of the variance is divided by the sixth reward value to obtain the corresponding advantage value.

9. The method for determining the parameters of an analog circuit according to claim 1, characterized in that, Determining the corresponding graph embedding vector based on the circuit source code includes: Convert the circuit source code into a netlist; The circuit diagram of the target analog circuit is constructed based on the netlist; The circuit diagram is input into the graph neural network model to obtain the corresponding graph embedding vector.

10. The method for determining the parameters of an analog circuit according to claim 9, characterized in that, The circuit diagram of the target analog circuit is constructed based on the netlist, including: Extract the netlist nodes and netlist edge information corresponding to the netlist, wherein the netlist nodes represent the components of the target analog circuit, and the netlist edge information represents the connection relationship between the components; The circuit diagram is obtained by processing the netlist nodes and netlist edge information using a graph data structure.

11. A parameter determination device for an analog circuit, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the parameter determination method for the analog circuit as described in any one of claims 1 to 10.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the parameter determination method for an analog circuit as described in any one of claims 1 to 10.

13. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the steps of the parameter determination method for the analog circuit according to any one of claims 1 to 10.

Images