Reconstruction of the Point Spread Function from Adonis Ao Measurements: Limits of the Method and Photometric Accuracy

{ Wavefront correction by adaptive optics is often only partial and the image of a point-like source consists of a coherent peak superposed on a diiuse halo. If the point-spread function (PSF) is known, the object can be restored by deconvolution techniques. This PSF depends, however, strongly on atmospheric conditions which are variable and on the wavefront sensor reference source. For this reason, the image of a point-like source does not always t well the PSF at the time of the object acquisition. It is thus important to estimate the PSF simultaneously with the observation of the object. Recently, we have reported on our rst results of PSF reconstruction on a bright reference source, observed with the adaptive optics system ADONIS. We now present results on faint reference sources observed under bad seeing conditions. We discuss the limits of the reconstruction method for both cases and nd that a variable residual aberration which we interpret in terms of a local turbulence can at least partly explain the error on the PSF reconstruction. From simulations, we studied the conservation of the ux ratio of double systems when using the reconstructed PSF as calibrator in deconvolution. It is shown that for ux ratios up to 2 the mean relative error on the ux ratio estimate is about 2-3%, while it is about 7% for ux ratios of 4 and 25% for higher ux ratios. The same errors are found, when using the image of a point-like source observed 30 min later as PSF calibrator.