Research On Bulk Acoustic Wave Filter Design Methods With Artificial Intelligence

With the rapid popularization of 5G technology in the world,the market demand for radio frequency filters continues to grow.Because of its advantages of high performance,low cost and small size,the bulk acoustic wave(BAW)filter has become the first choice for mobile radio frequency filters.However,the current BAW filter design process has the disadvantages of high time cost and dependence on the experience of engineers.Aiming at the shortcomings of BAW filter design,this thesis introduces AI algorithm to optimize BAW filter design.A prediction model of structural parameter range of BAW filter based on deep learning is proposed,which can reduce the design range of structural parameter of BAW filter.Through the Mason model of the programmable BAW resonator(BAWR)and the transmission matrix of the BAW filter,a large number of BAW filter data used to train the deep learning model are obtained in a short time.Then,according to the characteristics of the BAW filter data,a onedimensional convolutional hybrid network model is trained to predict the design range of the structural parameters of the BAW filter.This model can reduce the design range by 40% on the basis of the initial structural parameter design range.Through comparative experiments,it is verified that the introduction of deep learning to narrow the design range of structural parameters can shorten the design time of traditional BAW filters to about 3 hours,and lay a foundation for the subsequent research on structural parameter retrieval algorithms of BAW filters.A BAW filter structure parameter retrieval method based on improved artificial bee colony algorithm is proposed,which can automatically and efficiently obtain the BAW filter structure parameter combination meeting the design index.By building the BAW filter performance evaluation model as the retrieval standard of the artificial bee colony algorithm,the artificial bee colony algorithm is used to automatically retrieve the BAW filter structure parameter combination that meets the design indicators,and the introduction of engineering experience further improves the algorithm design efficiency and retrieval success rate.Finally,the effectiveness of the structure parameter retrieval method is verified through three groups of comparative experiments,This method can retrieve the BAW filter structure parameter combination that meets the design index with 90% success rate within 2 minutes.A BAW filter automatic placement algorithm based on genetic algorithm is proposed,which can realize the automatic placement of BAW filters and reduce the size of BAW filters.By introducing the concept of circular container,the placement difficulty of BAW filter is simplified,and the placement algorithm is suitable for BAWR with arbitrary shape.The success rate of the layout algorithm is improved through the greedy strategy,and the filling ratio of the layout is further improved through the compression strategy.Finally,the effectiveness of the placement algorithm is verified by two groups of BAW filter placement data.The placement algorithm can achieve BAW filter placement with a fill ratio of about 70%.