Numerical methods for SAW propagation characterization

Due to more and more stringent requirements on SAW filter performance, it is mandatory to precisely characterize the SAW propagation characteristics as a function of manufacturing variations (metal thickness, mark-to-space ratio, etc.). Several authors have already proposed experimental characterizations using sets of test devices. One of the main difficulties of this experimental approach is the accuracy of both the geometrical and electrical measurements. On the other hand, precise numerical methods have been developed to advantageously replace experiments. Generally speaking, these methods are CPU time consuming. However, due to the continuing improvement of computer power, they are becoming more and more time efficient when applied to the analysis of SAW filters. The overall efficiency depends on the numerical integration methods used. In this paper we present a review of the numerical programs that have been developed during the past few years. For both infinite and finite grating modeling, we developed numerical mixed FEM/BEM (Finite Element Method-Boundary Element Method) models using an efficient interpolation function basis that takes into account the singularity at both edges of each electrode. First we propose a model for the simulation of finite transducers with arbitrary geometries. This numerical code has been successfully used to analyze a SPUDT (Single Phase UniDirectional Transducer) on the Y+36° cut of LiTaO₃. For an infinite periodic grating, it is convenient to solve the propagation problem in the Fourier domain (wave number space and harmonic excitation) and important efforts have been spent to properly integrate the so-called periodic harmonic Green´s function. Using this numerical model together with the general P-matrix formalism, it is possible to compute all the basic parameters with a very good accuracy: these consist of the single strip reflectivity, acoustic wave-phase velocity, and phase offset between reflection and transduction centers. Simulations and comparisons with experiments are shown for each model