A Comparative Study of RF MEMS Switch Test Methods

Abstract

Micro-ElectroMechanical Systems, often abbreviated as MEMS are miniaturized elements, multiplied components and microelectronics that can capture and act on the environment in order to perform a number of missions. They have a small size, limited energy capacity and low memory capacity. MEMS applications exceed the scaling limits of current computational paradigms, posing serious challenges and new opportunities for information technology. The heart of the MEMS switches is a moving electrode that, by contacting a fixed electrode, creates modifications in an RF circuit. The movable electrode is often formed of a suspended beam exerting movement. Another form of the moving electrode is a thin disk suspended above the electrode system disposed on the substrate. The movement of the moving electrode is generated by an actuating force that is often electrostatic, but it can be thermal, piezoelectric, or magnetic. The purpose of the test is to discriminate between the good devices that respect its specifications and the faulty ones that are not functional. In general, the failures occur due to deviations in the parameters of the manufacturing process or the presence of manufacturing defects. This problem cannot be solved by a single method, but requires several complementary techniques. This paper will present a comparison between the test techniques of RF MEMS Switches. Thus, the proposed method is based on machine learning to create predictive models for testing devices. This reliable test strategy helps to overcome the necessity of sophisticated test equipment, as well as the access difficulties to measure embedded points.