Reducing Offset and Bias Instability in Lorentz Force Magnetic Sensors Through Bias Chopping

In this paper, we describe a method for reducing offset and offset-related 1/f (flicker) noise in micromechanical resonator-based Lorentz force magnetometers by chopping the dc bias voltage applied to the resonator. This 1/f noise is shown to determine the stability of the sensor's output at long averaging times. The magnetometer is operated at a 40-Hz frequency offset from its 62.88-kHz resonant frequency. At 4.6-mA bias current, the input referred noise is 70 nT/√Hz limited by white electronic noise. Using the proposed bias chopping method, the sensor's offset is reduced from 310 to 15 μT, and the bias instability is reduced from 27 to 7 nT. The averaging time to reach the bias instability increases from 18 to 290 s as a result of chopping. The root cause of 1/f noise is demonstrated to be 1/f noise on the ac and dc bias voltages applied to the sensor. While the sensor's readout is modulated at the 62.92-kHz excitation frequency and could therefore be expected to be free from 1/f noise, the baseband 1/f noise on these voltage sources is shown to be up-converted to the excitation frequency. Up-conversion occurs through the existence of fabrication imperfections that lead to mismatches in the sensing capacitances, resistances, and parasitic feedthrough capacitances of the sensor. By chopping the bias voltage, the Lorentz force signal is separated from the offset and 1/f noise produced by these error sources. This technique is a generic method to reduce 1/f noise and bias instability in sensors based on micromechanical resonators. [2016-0184].