From Adaptive Optics systems to Point Spread Function Reconstruction and Deconvolution for Extremely Large Telescopes
Speaker: Roland Wagner (RICAM)
Date: September 20, 2017 13:00
Location: SP2 416-2
Modern ground based telescopes like the planned European Extremely Large Telescope (E-ELT) depend heavily on Adaptive Optics (AO) systems, which use measurements of incoming wavefronts from guide stars to reconstruct the turbulence above the telescope and derive the shape of deformable mirror(s) (DM). The main challenge is to have a fast enough algorithm for solving an inverse problem arising in this process as the atmospheric turbulence is constant for approximately only 1 ms. Even though AO correction is used, the quality of astronomical images still is degraded due to the time delay stemming from the wavefront sensor (WFS) integration time and adjustment of the deformable mirror(s) (DM). This results in a blur which can be mathematically described by a convolution of the original image with the point spread function (PSF). The PSF of an astronomical image varies with the position in the observed field, which is a crucial aspect on ELTs. This talk gives an overview over different methods developed in my PhD thesis for the control of modern AO systems. Due to the provoking time constraints and the increased data sizes, the development of new methods is necessary. The methods presented yield a significant speed-up compared to standard AO reconstruction methods. The algorithms have been implemented in a state-of-the-art AO simulation environment and simulation results show that our methods obtain comparable quality while reducing the computational time significantly compared to established methods. For the post-processing, we present an approach for PSF reconstruction from WFS data combining atmospheric tomography and techniques for PSF reconstruction. Existing, verified techniques are fused together in a novel way to deliver accurate field dependent PSFs in very short time. To enhance the quality of the observed images with the PSF reconstructed from AO data, we opt for a blind deconvolution scheme to also improve the quality of the estimate for PSF. The choice of an efficient representation is crucial to get a sparse system.