(paging rhl )At least for my generated PSF the difference between calculating the center of the light distribution on the 10x oversampled images and the images in the native (detector) resolution is, at largest, around 2*10**-4 pixels in a single dimension. When I say pixel I am referring to the final pixels (15 microns.) This is shown on the image ``Difference_in_center_of_light''. I have placed values next to an arrow with one of the larger displacement values. 

I have then tested that the values are not overly sensitive to where the downsampling is happening, i.e., if the center of light at the center of the final pixel or at the edges. This is shown on the image ``single_explicit_example'', where we can see that no matter where the center of the light is, the difference between the oversampled image and the image in the native resolution is on the order of 10**-4 (shown as the title above the images on the right hand side).