| Micromechanical resonant devices have recently emerged as potential replacements for off-chip, discrete, frequency selective components, such as quartz resonators and SAW filters, in wireless communications systems. Microresonators maintain the high, mechanically-based quality factor of discrete components, but offer potential for integration with circuits, producing a fully-integrated, single-chip transceiver. Large-scale integration of microresonators, in numbers impractical for discrete parts, may enable new transceiver architectures with potential to greatly improve both power consumption and performance.;To meet the needs of a complete wireless system, microresonators must be scaled to function at frequencies ranging from the relatively low-frequency IF all the way to the RF carrier frequencies in products such as cellular phones. With such frequencies in mind, this work introduces a new micromechanical resonator design based on the radial-contour-mode vibration of a disk---an extensional mode that exhibits higher stiffness, and therefore higher frequency, than previous microresonator designs.;In the fundamental mode, disk resonators have been measured at frequencies as high as 433 MHz with quality factors of over 4,000 and with Q's in excess of 20,000 at 193 MHz. By utilizing higher vibration overtones, the resonant frequency may be scaled without the need to scale device dimensions. In the third overtone, disk resonators have been measured as high as 829 MHz, placing them well within the first U.S. cellular frequency band, and poised to break the 1GHz barrier.;Individual resonators have application as reference elements in oscillators, but the system-level benefits of micromechanical components are truly manifested by coupling multiple resonators to form filter networks. Basic filter operation has been demonstrated by mechanically coupling two disk resonators with an extensional coupling spring, yielding a two-peak spectrum characteristic of a two-resonator filter. Unfortunately, process limitations prevent full termination, making flat passbands unachievable, but the measured results agree very well with calculations, verifying the design theory. Further refinements in the process technology should produce improvements in measured results, leading to narrow-band, low-loss filter designs with even better performance than their discrete counterparts. |
