Convolution - weird effect - and high res fwhm (Mathcad15)

Hi. I am trying to determine the width that a 1.1um object would appear when imaged through a microscope and detected onto a camera.    This involves the convolution of 3 functions - the psf which quantifies the optics performance, the bead object itself - modelled as a top hat, and the detector pixel response again modelled as a top hat.

I cant understand why the artefact appears when I have a span of 100 but doesn't when the span is 80.

Also I perform the fwhm calc (via gaussian fit) on the pixellated image.  How would I more accurately perform it on the high resolution image and express the result in um.