Logo image
機械耦合式CMOS-MEMS濾波器之設計與特性探討
Thesis

機械耦合式CMOS-MEMS濾波器之設計與特性探討

陳昭瑜
Masters, 國立清華大學, 奈米工程與微系統研究所
2012

Abstract

金氧半導體微機電系統 機械耦合 電容式元件 微機械共振器 微機械濾波器 窄頻應用 阻抗匹配 雙埠量測 差分驅動
This work reports on the design and characterization of a mechanically-coupled CMOS-MEMS filter centered at several MHz with a narrow bandwidth and reasonable insertion loss after filter termination performed in a 4-port network analyzer. To implement a bandpass filter, the proposed filter structure produces two physical resonance modes, therefore forming a filter passband with a desired bandwidth by the use of the mechanical coupler at proper coupling locations. By the use of a conventional filter design, the metal stacking layers in a foundry-oriented CMOS platform were used to fabricate the device structures, making great progress aligned with existing fabrication lines. To further reduce the motional impedance, the high-velocity coupled array and gap-reduction mechanism were adopted to create larger transduction areas and tiny gap spacing for capacitive transducers, respectively. However, due to the existence of undesired parasitics from a typical two-port configuration, the resonance response would be dwarfed and masked by the background feedthrough floor. Furthermore, the matching condition is also limited by shunt capacitance at the I/O ports for filter transmission, that significantly barricades the proper filter termination. To solve this issue, a pure motional response of the proposed filter can be extracted once the de-embedding scheme is carried out. To attain the lower material loss, the silicon dioxide structure was developed in this work as well. Ideally, the SiO2 can provide better electromechanical coupling and higher quality factor as compared with the metal stacking counterparts. Based on the electrical isolation of the oxide structure, the filter devices can be operated in differential configuration via the balun function on the network analyzer. With that, the feedthrough parasitic can be evidently alleviated without any post-data processing, thus creating 30dB noise-floor improvement in filter performance.

Metrics

1 Record Views

Details

Logo image