SQP algorithm-based circuit parameter optimization method

Abstract

The invention provides a circuit parameter optimization method based on an SQP algorithm. The method comprises the steps that S1, a vector x is generated through parameters of a device of a to-be-optimized circuit; s2, defining a vector y through a constraint condition of a to-be-optimized circuit and a target performance index; s3, sampling is carried out in the sample space (X, Y) through a sampling algorithm, circuit simulation software is called to obtain a simulation result, and parameter pairs (x, y) obtained through sampling form a training set; s4, constructing a neural network model by fitting a function relationship of y = f (x); and S5, optimizing the neural network model generated in the step S4 through an order quadratic programming algorithm to obtain an optimal solution. According to the method, the simulation model is constructed by using the neural network, so that the distribution of the sample space can be described more accurately and comprehensively under the condition that the number of times of calling circuit simulation software is less. And the optimal circuit parameter combination under the structure can be obtained from a global perspective. The SQP algorithm is used to easily obtain a globally optimal solution and get rid of convergence of a locally optimal solution.

Description

technical field [0001] The invention relates to the field of integrated circuits, in particular to a circuit parameter optimization method based on an SQP algorithm. Background technique [0002] In the context of the continuous development of integrated circuits, all electronic devices require the participation of analog integrated circuits. Even the design of digital circuits requires analog modules to connect with the outside world. Our demand for analog circuit design is huge. In the process of analog integrated circuit design, it is necessary to weigh many kinds of performance indicators, and to find a balance point between multiple possible conflicting goals. It is necessary to continuously simulate, tune and improve through circuit simulation. At present, the existing technology uses genetic algorithms to simulate integrated circuits. The parameter optimization of the circuit has the disadvantages of slow convergence speed and many control parameters. [0003] In mos...

AI Technical Summary

Problems solved by technology

In the process of analog integrated circuit design, it is necessary to weigh many kinds of performance indicators, and to find a balance point between multiple possible conflicting goals. It is necessary to continuously simulate, tune and improve through circuit simulation. At present, the existing technology uses genetic algorithms to simulate integrated circuits. The parameter optimization of the circuit has the disadvantages of slow convergence speed and many control parameters.

[0003] In most cases now, when the design requirements and the constraint documents provided by the foundry are confirmed, we can determine the process based on this information, and every time a new structure is designed, the parameters of various devices in the circuit need to be reselected, manually It takes a lot of time to optimize circuit parameters, and the parameter optimization method of existing circuit simulation software is a scanning method, which can only be determined by traversing all possibilities for a specific parameter when other parameters are determined. In the optimal case, this method takes a long time and it is difficult to find the optimal parameter combination from a global perspective

Images