CN114638188B - An automatic parameter adjustment system and method for pre-simulation process of operational amplifier design - Google Patents
An automatic parameter adjustment system and method for pre-simulation process of operational amplifier design Download PDFInfo
- Publication number
- CN114638188B CN114638188B CN202210373866.1A CN202210373866A CN114638188B CN 114638188 B CN114638188 B CN 114638188B CN 202210373866 A CN202210373866 A CN 202210373866A CN 114638188 B CN114638188 B CN 114638188B
- Authority
- CN
- China
- Prior art keywords
- simulation test
- loop gain
- operational amplifier
- schematic diagram
- gain value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/36—Circuit design at the analogue level
- G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Amplifiers (AREA)
Abstract
The application relates to an automatic parameter adjusting system and method for an operation amplifier pre-design simulation process, wherein the system comprises a drawing input module, a simulation test module, a reinforcement learning module and an output storage module, wherein the drawing input module is used for drawing an operation amplifier schematic diagram, the simulation test module is used for reading the drawn operation amplifier schematic diagram with an open-loop gain value as a target, performing simulation test and outputting simulation test results, the reinforcement learning module is used for reading operation amplifier performance parameters obtained by the simulation test, automatically adjusting parameter values of devices needing parameter adjustment in the operation amplifier, controlling loop iteration test of the simulation test module, comparing and judging with the open-loop gain value until the open-loop gain value is met, and the output storage module is used for storing the simulation test results and the operation amplifier schematic diagram after parameter adjustment.
Description
Technical Field
The application relates to the field of analog integrated circuit design, in particular to an automatic parameter adjusting system and method for an analog process before an operational amplifier is designed.
Background
An operational amplifier is a circuit unit with a very high amplification factor, and in an actual circuit, a certain functional module is usually formed together with a feedback network. The operational amplifier design process has high demands on performance and efficiency, and thus, high demands are made on the design experience of the designer. In the previous simulation process of the operational amplifier design, a designer is required to continuously simulate the operational amplifier schematic diagram and adjust the parameter values of all devices in the schematic diagram in as short a time as possible until all performance parameters simulated before the operational amplifier meet the open-loop gain value of the operational amplifier design. The design process is extremely time-consuming and labor-consuming.
The traditional operational amplifier design method needs to carry out simulation test after the operational amplifier schematic diagram is drawn, so as to obtain the current performance parameters of the operational amplifier, then manually analyze and adjust the parameter values of each device according to the simulation result, and then carry out simulation test on the schematic diagram again so as to verify whether the manually modified schematic diagram can meet the requirement of the design open-loop gain value of the operational amplifier. The above process is repeated until more desirable parameter values for each device are sought.
The traditional method for performing the front-imitation parametric tuning of the operational amplifier design requires a large amount of preparation work before simulation in software, consumes great effort, and is unfavorable for the efficient implementation of the operational amplifier design flow because of low working efficiency of manual operation, undefined directivity and repeated parametric tuning and slow development speed in actual operation due to the complexity of verification.
Disclosure of Invention
In order to solve the problems of high effort, low efficiency and low design flow speed consumed by manual operation in the prior art that the design of the operational amplifier needs simulation test, the application provides an automatic parameter adjusting system for an imitation process before the design of the operational amplifier, which comprises a drawing input module, a simulation test module, a reinforcement learning module and an output storage module;
the simulation test module is used for reading the drawn operational amplifier schematic diagram with the open-loop gain value as a target, performing simulation test and outputting a simulation test result, wherein the simulation test result is an operational amplifier performance parameter, and the operational amplifier performance parameter corresponds to the open-loop gain value;
the reinforcement learning module is used for reading the simulation test result and judging whether the simulation test result accords with an open loop gain value or not;
if the simulation test result completely accords with the open-loop gain value, outputting the simulation test result and the schematic diagram to the output storage module, wherein the storage module is used for storing the simulation test result and the operational amplifier schematic diagram after parameter adjustment;
If the simulation test result does not accord with the open loop gain value, the reinforcement learning module adds a punishment value to the simulation test result according to a zero-order optimization algorithm, a covariance matrix self-adaptive evolution strategy algorithm, a Bayesian optimization algorithm and an expert priori algorithm, automatically generates new device parameters of the operational amplifier, modifies the device parameters in the schematic diagram, and outputs the modified schematic diagram to the simulation test module for simulation test again. And repeating the above process circularly until the simulation test result accords with the open loop gain value, and storing the schematic diagram and the simulation test result.
Further, the expert priori algorithm comprises a penalty function, and the penalty function is obtained by calculating an open loop gain value and operational amplifier performance parameters obtained through simulation test.
Further, the penalty function is specifically:
wherein f is the return value of the penalty function, f i is one term in the penalty function summation formula, N is the number of open loop gain values, x i is each open loop gain value, Is a performance parameter corresponding to each open loop gain value.
Further, the penalty function is:
Wherein f i is one term in the penalty function summation formula, x i is the open loop gain value with the smallest value in the penalty function, Is the corresponding performance parameter of the open loop gain value.
Further, the operational amplifier schematic diagram also comprises a component graphic symbol, a component character symbol, a component netlist, a component coordinate, a component rotation angle, a coordinate of each pin of the component, a connection line, a component position number coordinate, a circuit node coordinate and a character annotation.
Further, the steps are all completed under the Linux computer system.
The automatic parameter adjusting method for the simulation process before the operational amplifier design is applied to the automatic parameter adjusting system for the simulation process before the operational amplifier design, and the automatic parameter adjusting method for the operational amplifier design comprises the following steps:
Drawing an operational amplifier schematic diagram, which is used for reading the drawn operational amplifier schematic diagram with the open-loop gain value as a target, performing simulation test and outputting a simulation test result;
The drawing input module is used for drawing an operational amplifier schematic diagram, and is used for reading the drawn operational amplifier schematic diagram with an open-loop gain value as a target, performing simulation test and outputting a simulation test result;
reading the simulation test result, judging whether the simulation test result accords with an open loop gain value,
If the simulation test result completely accords with the open-loop gain value, outputting the simulation test result and the schematic diagram to the output storage module, wherein the storage module is used for storing the simulation test result and the operational amplifier schematic diagram after parameter adjustment;
If the simulation test result does not accord with the open loop gain value, the reinforcement learning module adds a punishment value to the simulation test result according to a zero-order optimization algorithm, a covariance matrix self-adaptive evolution strategy algorithm, a Bayesian optimization algorithm and an expert priori algorithm, automatically generates new device parameters of the operational amplifier, modifies the device parameters in the schematic diagram, and outputs the modified schematic diagram to the simulation test module for simulation test again. And repeating the above process circularly until the simulation test result accords with the open loop gain value, and storing the schematic diagram and the simulation test result.
A computer device, comprising:
a memory for storing a computer program;
and the processor is used for realizing the steps of the automatic parameter adjusting method based on the simulation process before the operational amplifier design when executing the computer program.
The technical scheme shows that the automatic parameter adjusting system for the simulation process before the operational amplifier design comprises a drawing input module, a simulation test module, a reinforcement learning module and an output storage module, wherein the drawing input module is used for drawing an operational amplifier schematic diagram, the simulation test module is used for reading the drawn operational amplifier schematic diagram with open-loop gain values as targets and conducting simulation test and outputting simulation test results, the simulation test results are operational amplifier performance parameters, the operational amplifier performance parameters correspond to the open-loop gain values, the reinforcement learning module is used for reading the simulation test results and judging whether the simulation test results accord with the open-loop gain values, if the simulation test results completely accord with the open-loop gain values, the simulation test results and the schematic diagram are output to the output storage module, the storage module is used for storing the simulation test results and the operational amplifier schematic diagram after parameter adjustment, if the simulation test results do not accord with the open-loop gain values, the reinforcement learning module is used for self-adapting to the algorithm according to a zero-order optimization algorithm, a Bayesian optimization strategy and a priori algorithm, and adding new operational amplifier parameters to the simulation test results, and revising the simulation parameters to the simulation test modules after the simulation test parameters are generated to the simulation test modules. And repeating the above process circularly until the simulation test result accords with the open loop gain value, and storing the schematic diagram and the simulation test result.
In the practical application process, the automatic parameter adjusting system and method for the simulation process before the design of the operational amplifier can automatically search and approximate ideal parameters of each device according to the initial parameter value and the open-loop gain value which are set manually based on the solution capability of reinforcement learning, can enable each device to find out the parameter value which meets the open-loop gain value, has better performance and lower power consumption in a shorter time, realizes automatic efficient design or auxiliary manual design, effectively integrates the processes of processing optimization of the simulation test flow, parameter adjustment of each device and the like, and improves the simulation efficiency before the design of the operational amplifier and the reliability of an optimization scheme.
Drawings
In order to more clearly illustrate the technical solution of the present application, the drawings that are needed in the embodiments will be briefly described below, and it will be obvious to those skilled in the art that other drawings can be obtained from these drawings without inventive effort.
FIG. 1 is a schematic diagram of an automatic parameter tuning system for an imitated process before designing an operational amplifier according to an embodiment of the present application;
fig. 2 is a schematic diagram of an automatic parameter adjusting method of an imitated process before designing an operational amplifier according to an embodiment of the application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more apparent, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments of the present application. The components of the embodiments of the present application generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
In the description of the present application, it should also be noted that the term "connected" should be interpreted broadly, unless explicitly stated or defined otherwise, and for example, may be an electrical connection or a communication connection. The specific meaning of the above terms in the present application will be understood in specific cases by those of ordinary skill in the art.
Furthermore, the terms "comprises," "comprising," and "including" and any variations thereof, are intended to cover a non-exclusive inclusion, such that a product or apparatus that comprises a list of elements is not necessarily limited to those elements expressly listed but may include other elements not expressly listed or inherent to such product or apparatus.
In order to solve the problems that simulation test is needed for the design of an operational amplifier in the prior art, labor consumption is high, efficiency is low and design flow speed is low, referring to fig. 1, a schematic diagram of an automatic parameter adjusting system for an imitation process before the design of the operational amplifier is provided in an embodiment of the application.
In the embodiment of the application, the drawing input module is used for drawing an operational amplifier schematic diagram, the simulation test module is used for reading the drawn operational amplifier schematic diagram with the open-loop gain value as a target, carrying out simulation test and outputting simulation test results, specifically, a user creates a file in the drawing input module, draws a nominated schematic diagram to be tested, edits parameter values, namely parameter values of components, such as W and L of a CMOS tube, which need to be adjusted in the schematic diagram, of a current source voltage source, setting simulation types, such as alternating current simulation, direct current simulation and transient simulation, designing the open-loop gain value, and then carrying out circuit simulation test. After the simulation is completed, the simulation test module generates a corresponding folder for storing simulation results in the folder, wherein the folder stores a schematic diagram file and a circuit netlist file generated by simulation, and simulation analysis results, and the simulation test module comprises information such as all static working points, all node voltages and currents and the like of alternating current simulation, direct current simulation and transient simulation analysis.
Further, in some embodiments of the present application, the reinforcement learning module is configured to read the simulation test result and determine whether the simulation test result meets an open loop gain value, where the open loop gain value is a performance index that needs to be met by a previous simulation result of the operational amplifier, and the performance index includes an open loop gain, a bandwidth, a phase margin, a slew rate, an output swing, an establishment time, an input impedance, an output impedance, noise, power consumption, a power supply rejection ratio, and a common mode rejection ratio.
In the embodiment of the application, if the simulation test result completely accords with the open-loop gain value, the simulation test result and the schematic diagram are output to the output storage module, the storage module is used for storing the simulation test result and the operational amplifier schematic diagram after parameter adjustment, namely the stored simulation test result and the operational amplifier schematic diagram are final results, and the reinforcement learning module receives the simulation test result obtained from the simulation test module when the operational amplifier design is optimized.
Further, in some embodiments of the present application, if the simulation test result does not conform to the open loop gain value, the reinforcement learning module automatically adjusts parameters of the schematic diagram, outputs the schematic diagram after automatic parameter adjustment to the simulation test module to perform the simulation test again, and repeats the above process in a circulating manner until the performance parameter of the operational amplifier obtained by the simulation test completely conforms to the open loop gain value, stores the schematic diagram and the simulation test result, and in the actual application process, a user needs to select the reinforcement learning algorithm and the iterative upper limit number for the reinforcement learning module in the automatic parameter adjustment system, manually set or select an initial value of each device parameter automatically generated by the expert experience algorithm, and set the open loop gain value of the operational amplifier schematic diagram design. The reinforcement learning algorithm comprises a zero-order optimization algorithm, a covariance matrix self-adaptive evolution strategy algorithm and a Bayesian optimization algorithm.
Further, in the embodiment of the application, the automatic parameter adjustment is realized by searching and generating different device specific parameter values according to a zero-order optimization algorithm, a covariance matrix self-adaptive evolution strategy algorithm and a Bayesian optimization algorithm, abstracting a simulation test module into a callable function program, inputting the function into the operational amplifier schematic diagram, outputting the function into the operational amplifier schematic diagram, calculating the simulation test result of the simulation test module into a specific floating point value according to a penalty function in an expert priori algorithm, continuously calling the operation simulation test module in a function call mode by a reinforcement learning module, inputting different device specific parameter values generated by the zero-order optimization algorithm, the covariance matrix self-adaptive evolution strategy algorithm and the Bayesian optimization algorithm into the simulation test module, and delivering the simulation test result to the penalty function, and re-searching and generating different device specific parameter values according to the generated device specific parameter values and corresponding values of the penalty function, so that the zero-order optimization algorithm, the covariance matrix self-adaptive evolution strategy algorithm and the Bayesian optimization algorithm can repeatedly find out the parameter values of the device specific parameter values until the performance of the device can meet the performance of the gain of the amplifier. And if the set iteration upper limit times are not found, forcibly ending the loop of the reinforcement learning module.
In some embodiments of the present application, the automatic parameter adjustment is specifically implemented in the following manner:
1. and establishing a sampling model D according to the initial values of the parameters of each device by a reinforcement learning algorithm.
2. The reinforcement learning algorithm samples and generates a plurality of different device specific parameter values from the model D, for example, if the element parameters to be optimized of the operational amplifier are W and L of a CMOS transistor in the circuit and a miller capacitance C, and the generated different parameter values are indicated by subscripts, the reinforcement learning algorithm generates a plurality of device specific parameter values (W1, L1, C1; W2, L2, C2; W3, L3, C3.; wm, lm, cm).
3. The reinforcement learning module continuously runs the simulation test module in a function call mode, inputs parameter values of various devices in different groups to obtain corresponding simulation test results, for example, if the simulation test of the operational amplifier is set to be open-loop Gain simulation, the Gain is used for representing a specific open-loop Gain value obtained by the simulation test result, and the subscript is used for representing the simulation test result of the parameter values in different groups, the simulation test results are obtained after the continuous simulation test (Gain 1, gain2, gain3, gainm).
4. According to the punishment function in the expert priori algorithm, the simulation test result of the simulation test module is calculated as a punishment value, which is a specific floating point number. For example, if the open-loop gain value GAINTARGET of the open-loop gain of the operational amplifier, the penalty function values (f 1, f2, f3,) of each group can be obtained by substituting GAINTARGET and each group of open-loop gain values mentioned in step 3 into the penalty function.
5. And constructing a new model D1 by the reinforcement learning algorithm according to the generated specific parameter values of the plurality of groups of devices and the penalty function values corresponding to the groups, and repeating the steps 4 above in a circulating way.
6. If a group of parameter values of each device which can enable the performance parameters of the operational amplifier to meet the open-loop gain value are found or the set iteration upper limit times are not found yet, ending the automatic parameter adjustment.
Further, in some embodiments of the present application, step 1, step 2, and step 5 in the automatic parameter adjustment implementation manner are automatically completed by a reinforcement learning algorithm.
In the embodiment of the application, if the simulation test result completely accords with the open-loop gain value, the simulation test result and the schematic diagram are output to the output storage module, the storage module is used for storing the simulation test result and the operational amplifier schematic diagram after parameter adjustment, namely the stored simulation test result and the operational amplifier schematic diagram are final results, and the reinforcement learning module receives the simulation test result obtained from the simulation test module when the operational amplifier design is optimized.
Further, in some embodiments of the present application, if the simulation test result does not conform to the open loop gain value, the reinforcement learning module performs preprocessing on the schematic diagram, outputs the preprocessed schematic diagram to the simulation test module to perform the simulation test again until the operational amplifier performance parameter obtained by the simulation test completely conforms to the open loop gain value, stores the schematic diagram and the simulation test result, in the actual application process, the user selects an iteration algorithm and iteration times in the automatic parameter adjustment system, sets the open loop gain value of the automatic parameter adjustment, that is, the target value required to be reached by each device parameter after parameter adjustment, the automatic parameter adjustment system can continuously and repeatedly perform the simulation test, process the schematic diagram simulation test result, automatically generate new operational amplifier device parameters by the selected reinforcement learning iteration algorithm until the operational amplifier performance parameter obtained by the simulation test result reaches the open loop gain value or the iteration times set by the user in the automatic parameter adjustment system, and outputs the final iteration result as a table for the user to select and store.
Further, in the embodiment of the present application, the preprocessing is specifically implemented by abstracting the simulation test module into a function according to a zero-order optimization algorithm, a covariance matrix adaptive evolution strategy algorithm, a bayesian optimization algorithm and an expert priori algorithm, the input of the function is a parameter value of each component in the operational amplifier schematic diagram, and the output of the function is calculated as a specific floating point value by a penalty function in the expert priori algorithm. The function is solved by a zero-order optimization algorithm, a covariance matrix self-adaptive evolution strategy algorithm and a Bayesian optimization algorithm, and parameter values of all devices which can enable the performance parameters of the operational amplifier to accord with an open-loop gain value are found out.
Specifically, in some embodiments of the present application, the reinforcement learning module performs preprocessing on a schematic diagram, and the reinforcement learning module reads initial parameter values of component parameters manually set by a user when the user performs initial operation of the simulation test module as input arguments of functions according to initialization parameters such as an iterative algorithm, iteration times, and search ranges of component parameters selected by the user, optimizes the entire schematic diagram, continuously searches and attempts new parameter values of component parameters, and performs repeated simulation test to implement iterative search, where a basic flow of the reinforcement learning module for solving the functions abstracted by the simulation test module is as follows:
establishing a sampling model D according to the initial values of the parameters of each device;
Sampling a set of solutions (e.g., x1, x2, x3,) from model D.
Calculating a corresponding function value f (xi) according to each solution xi;
constructing a new model D according to the acquired solution set and the corresponding function value;
according to the new model D, re-acquiring the solution set and the function value corresponding to the solution set until the convergence condition of the iterative algorithm is met;
the obtained optimal solution, i.e., each device parameter value enabling the performance parameter of the operational amplifier to conform to the open-loop gain value, is output.
In the embodiment of the application, the adopted zero-order optimization algorithm converts a constrained optimization problem into an unconstrained optimization problem by approaching an objective function or adding a penalty function to the objective function, the zero-order optimization algorithm does not utilize first derivative information, so that the time required by the zero-order optimization algorithm is less, the adopted covariance matrix self-adaptive evolution strategy algorithm is used for achieving the optimization purpose by simulating a natural biological evolution process, the algorithm has the characteristics of good global performance and high optimizing efficiency, a new path is provided for solving the problem of designing the optimization of the operational amplifier with high computing cost, the adopted Bayesian optimization is used for solving the extremum problem of the function with unknown expression, an acquisition function is constructed according to the regression result of the Gaussian process of the algorithm, and the extremum of the acquisition function is solved, so that the next sampling point is determined, so that the extremum of the function is obtained. All three reinforcement learning iterative algorithms adopt an open source scheme.
In some embodiments of the present application, an expert priori algorithm is introduced, which can help the iterative algorithm to search more quickly and effectively, so that the performance parameter of the operational amplifier accords with each device parameter value of the open-loop gain value, the initial value of the tuning parameter, that is, the parameter value set by the user in the simulation test, is suitable, which is helpful for better tuning, while the search range of each device parameter value, that is, the search range of the reinforcement learning algorithm for searching the target parameter value for each device parameter, is suitable, which is helpful for realizing balance between the tuning time and the tuning result, the penalty function penalty value calculation formula of the performance parameter obtained by each tuning simulation test, the return value of the penalty function is obtained by calculating the open-loop gain value and the performance parameter of the operational amplifier obtained by the simulation test, and the suitable penalty function value calculation formula is helpful for reinforcement learning to better converge.
In some embodiments of the present application, reinforcement learning is developed from theories such as learning from animals, adaptive control of parameter disturbance, etc., and the basic principle is that if a certain behavior strategy of a characteristic of a software and hardware system results in a positive prize for an environment, then a trend of the behavior strategy generated after the characteristic of the software and hardware system will be reinforced, the characteristic of the software and hardware system aims to find an optimal strategy in each discrete state to maximize a desired discount prize and maximum, reinforcement learning regards learning as a heuristic evaluation process, and the characteristic of the software and hardware system selects an action for the environment, and the state changes after the environment receives the action, and generates a reinforcement signal, such as a prize or a probability increase of punishment, and the selected action affects not only an immediate reinforcement value but also a state at a moment in the environment and a final reinforcement value.
According to the technical scheme, the automatic parameter adjusting system for the simulation process before the operational amplifier design comprises a drawing input module, a simulation test module, a reinforcement learning module and an output storage module, wherein the drawing input module is used for drawing an operational amplifier schematic diagram, the simulation test module is used for taking an open-loop gain value as a target, reading the drawn operational amplifier schematic diagram, conducting simulation test and outputting simulation test results, the simulation test results are operational amplifier performance parameters, the operational amplifier performance parameters correspond to the open-loop gain value, the reinforcement learning module is used for reading the simulation test results, judging whether the simulation test results accord with the open-loop gain value or not, outputting the simulation test results and the schematic diagram to the output storage module if the simulation test results completely accord with the open-loop gain value, and the reinforcement learning module is used for storing the simulation test results and the operational amplifier schematic diagram after parameter adjustment, if the simulation test results do not accord with the open-loop gain value, conducting self-adaption algorithm, bayesian optimization algorithm and expert matrix self-adaption algorithm are used for generating new simulation test parameters to the simulation test devices after the simulation test results are revised to the output the simulation device schematic diagram. And repeating the above process circularly until the simulation test result accords with the open loop gain value, and storing the schematic diagram and the simulation test result.
In the practical application process, the automatic parameter adjusting system and method for the simulation process before the design of the operational amplifier can automatically search and approximate ideal parameters of each device according to the initial parameter value and the open-loop gain value which are set manually based on the solution capability of reinforcement learning, can enable each device to find out the parameter value which meets the open-loop gain value, has better performance and lower power consumption in a shorter time, realizes automatic efficient design or auxiliary manual design, effectively integrates the processes of processing optimization of the simulation test flow, parameter adjustment of each device and the like, and improves the simulation efficiency before the design of the operational amplifier and the reliability of an optimization scheme.
In some embodiments of the present application, the automatic parameter adjustment system reads the initial value of each component parameter set by a user when performing a primary simulation test, and multiplies the initial value of each component parameter by a different coefficient to obtain a recommended parameter adjustment value search range of the parameter according to the common parameter search range of each component parameter of each different value. The search range is set to [7.5,12.5] when the element parameter having an initial value of [0, 50) is multiplied by a coefficient of 0.5 as the search range, for example, when the capacitance having an initial value of 10pF is used, the search range is set to [60,140] when the element parameter having an initial value of [50, 1000) is multiplied by a coefficient of 0.8 as the search range, for example, when the MOS transistor initial value W is 100, the search range is set to [1000,3000] when the element parameter having an initial value of 1000 or more is multiplied by a coefficient of 1 as the search range, for example, when the MOS transistor initial value W is 2000. And meanwhile, the user can modify the searching range of the tuning parameters of each parameter according to the own requirement.
In some embodiments of the present application, the expert a priori algorithm includes a penalty function calculated from the open loop gain value and the current operational amplifier performance parameters, typically using the same penalty function curve for each performance parameter when dealing with multiple open loop gain values.
Further, in some embodiments of the present application, the penalty function is specifically:
wherein f is the return value of the penalty function, f i is one term in the penalty function summation formula, N is the number of open loop gain values, x i is each open loop gain value, Is a performance parameter corresponding to each open loop gain value.
In some embodiments of the present application, the reinforcement learning module uses different penalty value calculation formulas for each performance parameter according to the open-loop gain value of the automatic tuning parameter set by the user and according to the expert priori algorithm and the numerical value size range of each open-loop gain value, so as to avoid that in the calculation process of the penalty function return value of the reinforcement learning algorithm, when each performance parameter differs from the corresponding open-loop gain value by the same proportion, the penalty value calculated by the performance parameter with a larger value is larger, and the penalty value calculated by the performance parameter with a smaller value is smaller, because the reinforcement learning algorithm will tend to find each component parameter value with a smaller penalty function return value, the final optimization result of the performance parameter corresponding to the open-loop gain value with a larger value may not conform to the set open-loop gain value.
Specifically, in some embodiments of the present application, if the open-loop gain value set by the user is 200dB open-loop gain, the bandwidth is 1MHz, and a reinforcement learning algorithm is operated, the program continuously searches for new parameter values of each component and uses the command line to invoke simulation to test the values of the performance parameters. Setting the parameter values of each component in one group to make the performance parameter value be the open loop gain 198dB and the bandwidth 1MHz, and setting the corresponding penalty function return value to be 4, and setting the parameter values of each component in the other group to make the performance parameter value be the open loop gain 200dB and the bandwidth 0.1MHz, and the corresponding penalty function return value to be 0.81. Reinforcement learning algorithms continually search for new component parameter values around component parameter values that make the return value of the penalty function smaller, i.e., find component parameter values that make the return value of the penalty function minimum.
In some embodiments of the present application, different penalty calculation formulas are used for each performance parameter, respectively, so as to adjust the influence of each open-loop gain value on the penalty function return calculation.
Specifically, in some embodiments of the present application, a penalty value calculation for each performance parameter is multiplied by a coefficient or a replacement penalty function calculation formula according to the magnitude of the value of each open loop gain value. For example, when the difference between the open-loop gain value with the largest value and the open-loop gain value with the smallest value is more than 40 times, the penalty value calculation formula corresponding to each open-loop gain value with the smallest value is multiplied by 1.0001, and when the difference between the open-loop gain value with the largest value and the open-loop gain value with the smallest value is more than 1000 times, the penalty value calculation formula corresponding to the open-loop gain value with the smallest value is replaced by the penalty function.
Further, in some embodiments of the present application, the open-loop gain value corresponding term with the smallest value in the penalty function is specifically:
wherein f i is a term of a penalty function summation formula, x i is an open loop gain value with the smallest value in the penalty function, Is the corresponding performance parameter of the open loop gain value.
In some embodiments of the present application, the operational amplifier schematic diagram further includes a component graphic symbol, a component text symbol, a component netlist, a component coordinate, a component rotation angle, a coordinate of each pin of the component, a connection line, a component bit number coordinate, a circuit node coordinate, and a text annotation.
In some embodiments of the present application, all of the above steps are performed under a Linux computer system.
In order to realize practical application of the system, the second aspect of the embodiment of the application also provides an automatic parameter adjusting method of an imitation process before the operational amplifier is designed, referring to fig. 2, which is a schematic diagram of the automatic parameter adjusting method of the imitation process before the operational amplifier is designed, the automatic parameter adjusting method of the operational amplifier is designed, and comprises the steps of drawing an operational amplifier schematic diagram, reading the drawn operational amplifier schematic diagram, performing simulation test and outputting simulation test results, the drawing input module is used for drawing the operational amplifier schematic diagram, reading the drawn operational amplifier schematic diagram, performing simulation test and outputting simulation test results, taking the open-loop gain value as the target, reading the drawn operational amplifier schematic diagram, reading the simulation test results, judging whether the simulation test results accord with the open-loop gain value, outputting the simulation test results and the schematic diagram to the output storage module if the simulation test results accord with the open-loop gain value completely, storing the simulation test results and the operational amplifier schematic diagram after parameter adjustment, and if the simulation test results do not accord with the open-loop gain value, reading the drawn operational amplifier schematic diagram, performing a self-optimizing algorithm, and revising the self-optimizing algorithm to the model algorithm according to the open-loop gain value, and optimizing the self-optimizing the algorithm, and correcting the self-adapting parameters, and generating the simulation parameter correcting algorithm, and the self-adapting device after the self-adapting algorithm. And repeating the above process circularly until the simulation test result accords with the open loop gain value, and storing the schematic diagram and the simulation test result.
A third aspect of some embodiments of the application provides a computer apparatus comprising:
and a memory for storing a computer program.
And the processor is used for realizing the steps of the automatic parameter adjusting method based on an operation amplifier pre-design imitation process when executing the computer program.
The technical scheme includes that the automatic parameter adjusting system for the simulation process before the operational amplifier design comprises a drawing input module, a simulation test module, a reinforcement learning module and an output storage module, wherein the drawing input module is used for drawing an operational amplifier schematic diagram, the simulation test module is used for reading the drawn operational amplifier schematic diagram with open-loop gain values as targets and conducting simulation test and outputting simulation test results, the simulation test results are operational amplifier performance parameters, the operational amplifier performance parameters correspond to the open-loop gain values, the reinforcement learning module is used for reading the simulation test results and judging whether the simulation test results accord with the open-loop gain values, if the simulation test results completely accord with the open-loop gain values, the simulation test results and the schematic diagram are output to the output storage module, the storage module is used for storing the simulation test results and the operational amplifier schematic diagram after parameter adjustment, if the simulation test results do not accord with the open-loop gain values, the reinforcement learning module is used for self-adapting to the algorithm according to zero-order optimization algorithm, optimization and expert matrix prior algorithm, the simulation test results are added to the simulation test results, and the simulation test parameters are revised to the simulation test devices after the simulation test results are generated, and the simulation device principle parameters are revised to the simulation test modules are output. And repeating the above process circularly until the simulation test result accords with the open loop gain value, and storing the schematic diagram and the simulation test result.
In the practical application process, the automatic parameter adjusting system and method for the simulation process before the design of the operational amplifier can automatically search and approximate ideal parameters of each device according to the initial parameter value and the open-loop gain value which are set manually based on the solution capability of reinforcement learning, can enable each device to find out the parameter value which meets the open-loop gain value, has better performance and lower power consumption in a shorter time, realizes automatic efficient design or auxiliary manual design, effectively integrates the processes of processing optimization of the simulation test flow, parameter adjustment of each device and the like, and improves the simulation efficiency before the design of the operational amplifier and the reliability of an optimization scheme.
The application has been described in detail in connection with the specific embodiments and exemplary examples thereof, but such description is not to be construed as limiting the application. It will be understood by those skilled in the art that various equivalent substitutions, modifications or improvements may be made to the technical solution of the present application and its embodiments without departing from the spirit and scope of the present application, and these fall within the scope of the present application.
Claims (6)
1. The automatic parameter adjusting system for the simulation process before the operational amplifier design is characterized by comprising a drawing input module, a simulation test module, a reinforcement learning module and an output storage module;
the simulation test module is used for reading the drawn operational amplifier schematic diagram with a preset open-loop gain value as a target, performing a simulation test and outputting a simulation test result, wherein the simulation test result is an operational amplifier performance parameter, and the operational amplifier performance parameter corresponds to the open-loop gain value;
The reinforcement learning module is used for reading the simulation test result and judging whether the simulation test result accords with an open loop gain value,
If the simulation test result completely accords with the open-loop gain value, outputting the simulation test result and the schematic diagram to the output storage module, wherein the storage module is used for storing the simulation test result and the operational amplifier schematic diagram after parameter adjustment;
If the simulation test result does not accord with the open loop gain value, the reinforcement learning module adds a penalty value to the simulation test result according to a zero-order optimization algorithm, a covariance matrix self-adaptive evolution strategy algorithm, a Bayesian optimization algorithm and an expert priori algorithm, automatically generates new device parameters of the operational amplifier, modifies the device parameters in the schematic diagram, outputs the modified schematic diagram to the simulation test module for carrying out simulation test again, and repeats the process circularly until the simulation test result accords with the open loop gain value, stores the schematic diagram and the simulation test result, and the expert priori algorithm comprises a penalty function, wherein the penalty function is obtained by calculating the open loop gain value and the operational amplifier performance parameters obtained by the simulation test;
the penalty function is specifically:
wherein f is the return value of the penalty function, f i is one term in the penalty function summation formula, N is the number of open loop gain values, x i is each open loop gain value, Is a performance parameter corresponding to each open loop gain value.
2. An automatic parametric tuning system for an operational amplifier pre-design simulation process according to claim 1, wherein the penalty function is:
Wherein f i is one term in the penalty function summation formula, x i is the open loop gain value with the smallest value in the penalty function, Is the corresponding performance parameter of the open loop gain value.
3. The system of claim 1, wherein the operational amplifier schematic diagram further comprises a component graphic symbol, a component text symbol, a component netlist, a component coordinate, a component rotation angle, a component pin coordinate, a connection line, a component bit number coordinate, a circuit node coordinate, and a text annotation.
4. The automatic parametric tuning system for an operational amplifier pre-design simulation process of claim 1, wherein the steps are performed under a Linux computer system.
5. An automatic parameter tuning method of an operation amplifier pre-design imitation process, which is characterized in that the automatic parameter tuning method is applied to an operation amplifier pre-design imitation process automatic parameter tuning system according to any one of claims 1-4, and the operation amplifier design automatic parameter tuning method comprises the following steps:
Drawing an operational amplifier schematic diagram, which is used for reading the drawn operational amplifier schematic diagram with the open-loop gain value as a target, performing simulation test and outputting a simulation test result;
reading the simulation test result, judging whether the simulation test result accords with an open loop gain value,
If the simulation test result completely accords with the open-loop gain value, outputting the simulation test result and the schematic diagram to the output storage module, wherein the storage module is used for storing the simulation test result and the operational amplifier schematic diagram after parameter adjustment;
if the simulation test result does not accord with the open loop gain value, the reinforcement learning module adds a punishment value to the simulation test result according to a zero-order optimization algorithm, a covariance matrix self-adaptive evolution strategy algorithm, a Bayesian optimization algorithm and an expert priori algorithm, automatically generates new device parameters of the operational amplifier, modifies the device parameters in the schematic diagram, outputs the modified schematic diagram to the simulation test module for carrying out simulation test again, and repeats the process circularly until the simulation test result accords with the open loop gain value, and stores the schematic diagram and the simulation test result.
6. A computer device, comprising:
a memory for storing a computer program;
A processor for implementing the steps of the automatic parameter tuning method based on an operational amplifier pre-design simulation process as claimed in claim 5 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210373866.1A CN114638188B (en) | 2022-04-11 | 2022-04-11 | An automatic parameter adjustment system and method for pre-simulation process of operational amplifier design |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210373866.1A CN114638188B (en) | 2022-04-11 | 2022-04-11 | An automatic parameter adjustment system and method for pre-simulation process of operational amplifier design |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114638188A CN114638188A (en) | 2022-06-17 |
| CN114638188B true CN114638188B (en) | 2025-05-23 |
Family
ID=81952360
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210373866.1A Active CN114638188B (en) | 2022-04-11 | 2022-04-11 | An automatic parameter adjustment system and method for pre-simulation process of operational amplifier design |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114638188B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116702665B (en) * | 2023-08-04 | 2023-10-27 | 山东启芯软件科技有限公司 | Large-scale integrated circuit design optimization method based on common sub-expression |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105989199A (en) * | 2015-01-29 | 2016-10-05 | 西安电子科技大学昆山创新研究院 | Simulation method and device of operational amplifier |
| CN108614904A (en) * | 2016-12-11 | 2018-10-02 | 复旦大学 | Fast Analysis Method of Integrated Circuit Parameter Yield Based on Simulation Data Correlation |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6625785B2 (en) * | 2000-04-19 | 2003-09-23 | Georgia Tech Research Corporation | Method for diagnosing process parameter variations from measurements in analog circuits |
-
2022
- 2022-04-11 CN CN202210373866.1A patent/CN114638188B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105989199A (en) * | 2015-01-29 | 2016-10-05 | 西安电子科技大学昆山创新研究院 | Simulation method and device of operational amplifier |
| CN108614904A (en) * | 2016-12-11 | 2018-10-02 | 复旦大学 | Fast Analysis Method of Integrated Circuit Parameter Yield Based on Simulation Data Correlation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114638188A (en) | 2022-06-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN118428492B (en) | Prompting word optimization method, prompting word optimization system, electronic equipment and storage medium | |
| Butti et al. | Model order reduction based power system stabilizer design using improved whale optimization algorithm | |
| CN113468792B (en) | Parameter correction method and device for electromagnetic transient simulation model and electronic equipment | |
| CN118312604B (en) | Natural language query method and device | |
| CN111881664B (en) | Information extraction method, device, equipment and medium combining RPA and AI | |
| CN114638188B (en) | An automatic parameter adjustment system and method for pre-simulation process of operational amplifier design | |
| CN119415661A (en) | Method, device, equipment, medium and program product for responding to user question information | |
| CN118862970B (en) | Large language model, model training method, device, medium, equipment and product | |
| CN118095365A (en) | Model conversion method, device, equipment, storage medium and product | |
| CN118194954A (en) | Neural network model training method and device, electronic device and storage medium | |
| CN111598294A (en) | Active power distribution network reconstruction algorithm and device based on improved teaching optimization | |
| CN110929087A (en) | Audio classification method and device, electronic equipment and storage medium | |
| US20220076107A1 (en) | Neural network processing device, data processing method and device | |
| CN104376120B (en) | A kind of information retrieval method and system | |
| CN112132281B (en) | Model training method, device, server and medium based on artificial intelligence | |
| CN112182058B (en) | Content acquisition method, device, computer equipment and medium combining RPA and AI | |
| US20230169240A1 (en) | Computing device and method generating optimal input data | |
| CN118536491A (en) | Low-resource unified information extraction method based on high-efficiency fine adjustment of large model parameters | |
| Hedlund | Electrical optimization of PLAs | |
| WO2024165123A1 (en) | Devices and methods for providing a transformer neural network model | |
| KR20240086955A (en) | Simulation apparatus for fish farm based on digital twin | |
| Xu et al. | An adaptive multi-objective multifactorial evolutionary algorithm based on mixture Gaussian distribution | |
| CN115940202A (en) | Multi-inverter power distribution control method, device and equipment based on artificial intelligence | |
| Xu et al. | Pluggable neural machine translation models via memory-augmented adapters | |
| CN119416822B (en) | Task processing method, system, terminal and medium based on multiple expert layers |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |