PSF reconstruction via full turbulence estimation and telescope end-to-end simulation

Abstract

Enhancement and wise archiving of astronomical images require an accurate estimate of the observational Point Spread Function (PSF). Although modelling of the telescope and its optics is a well-understood problem, PSF reconstruction becomes challenging when the observations include adaptive optics (AO) correction. The approach presented in this paper consists in feeding an end-to-end (E2E) simulation of the telescope, the instrument and its environment with the characterized disturbances from the telemetry and AO loop data, in order to produce estimated PSFs.This method benefits from the developments made in the last years with respect to the estimation of external disturbances during AO correction, such as turbulence profile and its dynamics as well as sensor noise and vibrations characteristics. Internal parameters including dead actuators or non-common path aberrations are also considered when provided by the calibration. Once identified, these internal and external parameters are used as inputs to carry out E2E simulations of the optical propagation and estimate the PSF. The advantages of such a “brute force” method for PSF reconstruction – E2E simulations are highly intensive in computer power for ELTs – reside in its ability to integrate complex combination of effects from all the disturbances. It avoids analytical approximations used in classical approaches for the aliasing or fitting errors. E2E simulations have been used before in PSF reconstruction, but limited to a theoretical modelling of the system. Here, the authors use powerful GPUs that can reduce the processing time to a couple of orders of magnitude with respect to the real-time case. The object-oriented Matlab adaptive optics package has been used for the simulations. We analyze how the real-time gap can be further reduced using dedicated hardware modules (e.g. FPGAs) for the most stable parts of the code such as turbulence realizations and WFS simulations.