Neural-Network-Based Wide-FOV CMOS Sun Sensor

This paper deals the operation and test of a micro digital sun sensor prototype for precise measurement of the sun line. The sensor concept relies on producing multiple sun images on the focal plane by means of an opaque mask with multiple apertures. This improves measurement precision thanks to the average of multiple simultaneous measures. Nevertheless, sensor operation with a wide field of view is more complex since it is has to be performed with a variable number of spots. This poses several problems, including setting up a suitable calibration function for operation with variable spots and the possibility of failures in the identification of spots near the FOV borders. Main focus of the present paper is on the characterization of the wide field of view, variable-spot operation mode of the sensor by means of laboratory tests carried out by an ad-hoc test facility, also described in the paper. Results demonstrate that, thanks to the use of neural calibration and the introduction of a new image acquisition procedure characterized by variable shutter time, it is possible to keep high measurement precision, in the order of 5 arcsec, also for large off-boresight angles.