The performance of the Smartt-WFS with machine learning in simulation and lab bench testing

Abstract

Wavefront sensing is an integral part of all adaptive optics systems as it allows for the phase of the incoming light to be measured and corrected. In strong scintillation conditions, such as during daytime, or horizontal propagation, traditional, slope-based wavefront sensing techniques have a large decrease in performance. We investigate the use of the Smartt wavefront sensor in strong scintillation. The performance of the Smartt wavefront sensor has been shown to be insensitive to scintillation, however, at the cost of a reduced linear range. To overcome the reduced linear range we have investigated the use of machine learning as a phase reconstructor, opposed to the traditional linear estimator. We present the performance of the Smartt wavefront sensor in end-to-end adaptive optics simulations with a machine learning reconstructor as well as preliminary results from lab bench testing.