Three dimensional integration and packaging using silicon micromachining

This thesis presents a variety of structures for high frequency applications utilizing silicon micromachining. A study of W and X-band circuits is presented providing results needed for the design and fabrication of high-isolation systems. FGC lines in comparison to microstrip have been shown to give 8 dB higher isolation. Interconnects printed on single or multiple layers and separated by a distance equal to a λ_d demonstrate an isolation of −50 or −60 dB for W and X-band respectively. Moreover, lines printed in close proximity to open-end discontinuities suffer a degradation in coupling by 6 dB. Furthermore, the micromachining of dielectric cavities has been studied and shown to provide an increase in isolation on the order of 5 dB. The existence of vias next to RF interconnects has been investigated as a worse case scenario and demonstrated reduced isolation by 9 dB. The possibility of achieving good RF performance on low resistivity silicon wafers with the use of intermediate layers is proposed. The measurements presented demonstrate the feasibility of such an approach. Evanescent mode filters on high resistivity silicon wafers have been included in this research. Such structures offer a Q of 300 and can be easily integrated with other planar architectures. Due to their size (5 x 5 mm), these resonators can be tuned using appropriate MEMS devices. Starting an investigation of possible tuning mechanisms for micromachined cavities led to the design of an on-wafer packaging scheme for RF MEMS switches utilizing silicon micromachining. This is the first time that such an on-wafer package (dimensions: 2 x 2 mm, height 40 μm) is presented. The outcome of this effort is a DC-to-40 GHz package with an insertion loss of 0.06 dB due to the via transitions. Since this is an on-wafer package no wire-bond or other type of interconnect is necessary for accessing the MEMS. An initial hermeticity test is performed using packaged dew-point sensors. The measurements show a mean time to failure of 577 hours under accelerated conditions (130°C, 2.5 atm, 100% RH), which relates to more than 200 hours for “tropical” conditions (35°C, 95% RH).