SAW Filter Frequency Prediction with Optical Metrology Data and Machine Learning

Abstract

The most critical parameters that determine the frequency of RF SAW filter devices are the CD and pitch of the interdigital transducer (IDT) finger. The frequency is also affected by the thickness of passivation dielectric layers and their mechanical properties such as sound velocity. In this paper, we are going to present how SAW filter frequency can be predicted with optical metrology data and machine learning (ML) approaches without electrical tests. The predicted frequency then can be used to improve process control and minimize yield loss by detecting process excursions earlier.