| SAW(Surface Acoustic Wave,SAW)ladder filter is an essential component of the RF frontend,and it is widely used in mobile communication devices such as mobile phones.However,ladder filters have problems such as complex design and multiple design freedom.This study introduces swarm intelligence algorithms and machine learning methods to improve the efficiency of filter design.The specific research content and results are as follows:(1)The COM(Coupling of Modes,COM)model of the SAW ladder filter is established.The COM equation and P matrix of SAW devices are deduced and programmed using MATLAB to calculate the electrical performance of resonators and filters quickly.A ladder filter is simulated using the constructed COM model,and the accuracy of the model is verified by comparing the simulation results with reference results.(2)A traditional ladder filter design software is developed to achieve an automatic and efficient design.Aiming at the problems of low search efficiency and a large number of iterations in simulated annealing algorithms commonly used in SAW ladder filters,a new swarm intelligence algorithm,the artificial bee colony algorithm,is introduced.The optimization results of the simulated annealing algorithms and artificial bee colony algorithm under the same optimization objective are compared,it is found that the artificial bee colony algorithm converges within 500 generations,with an average insertion loss of 1 d B,while the convergence algebra of the simulated annealing algorithm exceeds 2000,with an average insertion loss is 1.2 d B.Compared with the simulated annealing algorithm,the artificial bee colony algorithm enhances the optimal value of the solution by 15% and improves the search efficiency by four times.In order to avoid the problem of excessively large solution space caused by total parameters optimization,a stepwise optimization strategy is adopted: first,the initial values of electrode thickness and interdigital period are determined based on empirical formulas and ergodic methods,and then the parameters such as interdigital logarithms,aperture,interdigital period,electrode thickness of each resonator are optimized by three steps using an artificial bee colony algorithm.(3)Based on the traditional ladder filter design software,an intelligent design method of a new type of SAW filter using an artificial bee colony algorithm combined with the BP(Back Propagation,BP)neural network and COM model is proposed.The efficient design of the new kind of SAW ladder filter is realized using this method.Aiming at the difficulty in extracting COM parameters of the new type of SAW filter(taking a thin film SAW filter as an example),machine learning is used to predict the COM parameters of the thin film SAW filter.Firstly,two newly introduced design parameters in thin-film SAW resonators are studied using the finite element method.It is found that they have a significant impact on the frequency and electromechanical coupling coefficient of the resonator,which in turn affects the central frequency and bandwidth of the filter.Through the finite element 2D unit model and orthogonal experiments,572 sets of training data are collected.Then a 13×13 BP neural network is trained,which can accurately predict the COM parameters of thin film SAW resonators with a prediction accuracy of 99%,avoiding the operation of repeatedly modeling and extracting parameters by finite element methods,significantly improving design efficiency.Finally,the design time of the thin-film SAW filter is shortened to 3 hours.(4)The electromagnetic parasitic effects under high-frequency conditions are introduced into the design software to achieve high-frequency expansion of the software.The COM model only considers the electro-acoustic characteristics of the filter without the electromagnetic parasitic effects of the bus bars at high frequencies.The electromagnetic simulation model of the SAW ladder filter is established in the high-frequency structure simulation software.The electromagnetic parasitic effect is equivalent to capacitance and inductance and is then introduced into the design software by converting it into a transmission matrix.(5)Case validation of intelligent design methods of SAW filter.The design of an interphone band ladder filter and a Band3 Tx ladder filter is completed using the developed design software,respectively.The designed filters are imported into the circuit software to analyze the power spectrum and time domain waveform of the input and output signals.It is found that the two filters attenuate the signal in the passband by 1.5 times and 1.12 times,respectively,and attenuate the signal in the stopband by more than 10000 times,which demonstrates good filtering effects.Finally,the automatic drawing program for resonator layout is written in MATLAB to generate a single resonator layout and then connect each resonator in ADS to complete the filter layout. |
PDF Full Text Request