Strategy for Determining a Magnet Position in a 2-D Space Using 1-D Sensors

Abstract

In this paper, we present a strategy employing four coplanar magnetic sensors with a single sensitive direction to detect the position of a magnet in a 2-D plane parallel and above the sensors’ plane. By positioning the magnet in such a way that its magnetic moment vector lies on the direction orthogonal to the sensing elements’ sensitive direction, the signal on the sensor plane assumes a quadrupolar characteristic, which takes advantage of both the negative and positive sensing ranges of the magnetic sensors, maximizing signal variability in the sensing plane. To validate this strategy, a sensing device with a $3\times 3$ mm2 active area containing four magnetic tunnel junction arrays as magnetic sensors was fabricated. Since the combination of the measured signal in each sensor is correlated with the magnet position, these were fitted using predictive methods to map both the quantities to each other. The two best performing predictive methods, artificial neural network fitting and regression forest fitting, were verified against an independently acquired data set, yielding a position root mean square error of 178 and $96~\mu \text{m}$ , respectively.