Harnessing Embedded Magnetic Fields for Angular Sensing With Nanodegree Accuracy

A multisensor approach that capitalizes on the existing magnetic fields in permanent-magnet-based actuators to achieve unobtrusive high-accuracy position sensing is presented. As magnetic-field models are position dependent, their inverse problems are often highly nonlinear with nonunique solution. This paper illustrates the principle and motivation for a multisensor approach using the concepts of parametric spaces to take advantage of multiple independent sensor measurements to induce a unique field-position correspondence in multisource fields. A direct mapping approach using supervised back-propagation artificial neural networks is utilized to attain positional information from distributed field measurements. Using an experimental rotary setup containing 24 magnetic sources, the measurements obtained from a network of magnetic Hall-effect sensors are statistically characterized and used to investigate the factors affecting the accuracy of the sensing system. Of particular interest are the combined effects of the number and spatial configuration of the sensors. Two types of sensor arrangement are investigated: an in-phase configuration consisting of evenly spaced sensors and a staggered configuration where unevenly spaced sensors concurrently measure different points of a periodic field. Using a network of 24 single-axis Hall-effect sensors in staggered configuration, the system is capable of achieving nanodegree angular positional accuracy.