Power amplifier design method based on feasible region shrinkage Bayesian optimization

Abstract

The invention discloses a feasible region shrinkage Bayesian optimization-based power amplifier design method, which comprises the following steps of: firstly, determining a matching network target impedance based on load traction and source traction, and then determining a matching network structure and element parameter values through a Chebyshev low-pass topology method; sampling an initial value by using optimal Latin hypercube sampling to obtain an input sample set, realizing sample point evaluation in a normalized weighting mode, searching an optimal hyper-parameter by using a particle swarm algorithm, maximizing a collection function UCB in a feasible region based on a trained Gaussian process model to obtain a next evaluation point, and continuously iterating to obtain an evaluation result; and the broadband high-efficiency power amplifier is realized. The invention provides a power amplifier design method based on feasible region shrinkage Bayesian optimization for the first time, and the feasible region is shrunk by dynamically changing the parameter value beta, so that the optimization is balanced between a global state and a local state, the convergence speed is increased, the power amplifier layout optimization can be effectively guided, and the power amplifier design is realized.

Description

technical field [0001] The invention belongs to the field of machine learning and radio frequency power amplifier design, and in particular relates to a power amplifier design method based on feasible domain contraction Bayesian optimization. Background technique [0002] The design indicators of power amplifiers, such as output power and efficiency, are mutually restrictive, and satisfying these indicators at the same time requires complex topology design and accurate parameter calculation, and even if the selected parameters conform to "Pareto optimal", once the manufactured The power amplifier does not match the analog design, and the designer must repeat the entire design cycle again, so the design process of the power amplifier is very complicated. The general steps include load pulling and source pulling, designing topology, ensuring circuit stability, component parameter calculation and circuit optimization, etc. The most important step is circuit optimization, and an...

AI Technical Summary

Problems solved by technology

[0003] The optimization of the power amplifier involves the length and width of the microstrip line of the matching network. There are many parameters, forming a huge high-dimensional parameter space, which makes the optimization problem face two difficulties. On the one hand, the unknown form of the objective function makes it impossible for the designer to use the gradient information, on the other hand, the calculation of the objective function has uncertainty, and the function value cannot be accurately estimated

On top of that, during the design process, the evaluation of the objective function takes a lot of time, money and human resources

In other words, power amplifier optimization is actually an expensive high-dimensional black-box optimization problem, and finding the optimal solution in high-dimensional space is a difficult problem, and the embedded optimization tools in electronic design automation (EDA) tools often have a long optimization time. , Problems such as error reporting and layout debugging difficulties during optimization

Images