Analysis of the noise vs. number of stacked images. A simple version of analyses I've seen in the past...I've just never documented my own experiments until now. This is a 300x crop of the Soul Nebula in 5nm Ha, QSI683wsg (Kodak 8300 @ -20C) on an FSQ-106 EDXiii under dark skies. Each frame is 1800 seconds, calibrated with 30 flats, 100 bias and 30 darks. PI's NoiseEvaluation Script was used to plot the noise measurements and produce the graph.
For consistency all images were integrated with ImageIntegration using Percentile Clipping, and all are shown with AutoSTF. Higher SNR can be achieved with WSC given a sufficient number of subframes (min 8 for WSC, for example).
Moral of the story, with this or similar cameras stack around 16 frames, calibrate properly and you've hit the asymptotic area of the noise reduction curve. Different cameras, conditions and calibration will give different results, but this is pretty consistent with things that I've only heretofore read and not done myself. Pure mathematicians may be able to state this more clearly, but I just wanted to prove it for myself. Remote access to consistently clear, dark and steady skies affords additional opportunities to test things and not worry too much about running from impending clouds that are always just over the horizon in my backyard....
I've found this material published by Dennis Isaacs to be very informative...much more so than my simplified analysis:
http://www.dens-astropics.org.uk/page%2035.htm
Comments