| | 7 | The next plot is where we disagree. This plot shows the mean pixel sky level divided by the number of input warps for the stack, as a function of the number of input warps. Number of input warps scales with exposure time of the final stack, so the x-axis on this plot is in large part just a proxy for the sky sigma of the image. The bias in the psphot background mean measurements has a sigma dependence, due to the way we perform the image histograms. I think a similar effect occurs no matter how the histograms are done; at some level the calculated value of the mean depends on how well the inner N sigma is sampled. In any case, a sigma dependent bias scales as teh square root of the number of input images, so the plotted value of the sky level divided by number of inputs produces a 1/sqrt(N) shape, which matches reasonably well with the values shown here. |
