Exploring the impact of PSF reconstruction errors on the reduction of astronomical adaptive optics based data

Abstract

ABSTRACT Astronomical adaptive optics (AO) data analysis requires the knowledge of the PSF associated to the AO run.With new AO modes soon to become available (LTAO, G LAO) there is a request from the AO users communityfor the development of new PSF reconstruction algorithms. Question is: what is the required accuracy on thequality of the reconstruction ? Guide-lines are needed in order to check the validity/usefulness of a given PSFreconstruction approach. LTAO/GLAO PSF reconstruction algorithms are being studied but are not availableyet, so we propose to analyze this issue by simulating AO data with an AO modeling tool, and studying the datareduction using modeled AO PSF with an increased level o f dierence with the initial AO PSF, for parametersthat have potentially a large impact on the PSF structure: the seeing angle, the C 2N vertical distribution,the residual tip-tilt, LGS altitude ”uctuations, and o- axis PSF variation (anisoplanatic eects). Results aregiven in the context of data analysis of an LTAO mode mimicking the planned VLT/GALACSI system. Wedo not take into account any instrument mode, and the telescope is assumed without aberrations. The currentstudy is focused on the most critical type of data reduction: deconvolution. Algorithms are reviewed, andit is shown that for most classical d econvolution methods, the main impact of PSF reconstruction errors canalready be described using either the so-called residua l “lter (ratio of exact OTF ov er reconstructed OTF) ormore simply the dierence between the exact and the reco nstructed PSF. Using these two metrics, we explorethe consequences of uncertainties on the “ve parameters introduced above. It is found that (1) in general, theimpact of PSF reconstruction errors, while noticeable, appears to be surprisingly low, relaxing apparently theneed for highly sophisticated PSF reconstruction algorithms; (2) in the case of Wiener-like deconvolution, tip-tiltuncertainty is the most critical parameter, and has a noticeable impact on the residual PSF wings - which canbe a problem when looking for faints objects in the vicinity of a bright star; (3) in the case of source extraction(CLEAN-like algorithms), seeing error clearly dominates, and the other errors have basically the same impact;(4) the impact of numerical eects during PSF deconvolution or extraction (sub-pixel PSF positioning error) isof the same order of magnitude than the eect of AO PSF parameters uncertainties.Keywords: Adaptive optics, PSF reconstruction, Image deconvolution