MEMS real-time clocks based on epitaxial polysilicon: system-level requirements and experimental characterization

The purpose of this paper is to assess the feasibility of MEMS-based real-time clocks (RTCs) using conventional polysilicon, without correcting the temperature coefficient of frequency (TCf) through dedicated technological steps. The paper first shows how such a large TCf (-30 ppm/K) is not an issue in terms of maximum frequency correction to achieve with a dedicated electronics: indeed, whatever the TCf, the dominant part of the frequency correction, required to match the 32-kHz RTC target value, is always demanded by the native frequency offset due to etching nonuniformities, and not by temperature changes. This sets the required number of bits of the modulator used to drive a fractional frequency divider that performs the compensation. Instead, requirements in the bit number and refresh rate of the temperature sensor are affected by a large TCf. Nevertheless, the paper shows the possibility to achieve few ppm frequency stability using a 9-bit temperature sensor with a 4-Hz refresh rate. This makes the approach quite competitive against more sophisticated MEMS processes, especially in terms of final cost. Experimental measurements on a MEMS-based resonator coupled to a dedicated integrated circuit are used to support the discussion.