Having recently purchased the ZWO ASI294mc, I was wondering how many darks I need to take so as not to be subtracting too much noise and end up subtracting too much random "false noise" from my images. On various forums, some say 30, others 40 or 50 or more. So I thought I'd try work it out for myself.
Various sets of darks are integrated all at 300", -10 degrees, 121 gain, and 30 offset, and graphed the output of noise using the NoiseEvaluation script in Pixinsight. I understand that every camera is different, so these results will only help asi294mc owners. I am also aware that the NoiseEvaluation script is not perfect, but only estimates noise based on extrapolation. Regardless though, I used a best fit exponential curve, and found it interesting that the curve is still notably decreasing after 30 and 40 dark frames. In fact, it might continue to decrease slightly even after 50. Not sure whether this would produce noticeable results though at such incremental decrease.
Based on these results, I will be taking no less than 50 darks for each set of imaging parameters (temperature, gain, offset, exposure time).
I hope you find this interesting, and feel free to let me know if I have made any mistakes in the analysis.
Thanks Sean, this is a great video with tons of interesting details! Although with his camera, the noise curve flattens much quicker than mine at 15. Very insightful though. He uses this method for his flats and dark flats too.
I do 30 darks. Works for me with an ASI294MC. The diminishing returns of adding more dark frames after that point makes taking more not very worthwhile for me. Taking more darks would improve my image, but probably not by any noticeable amount. Perhaps a single extra sub would be the difference in Signal to noise between 30 and 40 darks.