Series 1: 100% crops of zero light images for 15 second exposures, 100, 200, 400, 800, 1600, 3200 ISO...

Series 2: Same images as Series 1, with 8x light amplification ( levels adjusted ) applied...

Series 3: Shot to shot noise. Identifying noisy pixels. I used the 100 and 400 ISO shots, made one into a B channel and one into R channel then combined a looked for purple. The purple pixels represent noisy pixels from one shot to the other, e.g. not random noise, but only the noisy pixels. It is important to keep in mind that based on the analysis below, much much more noise is introduced by random effects and the small number of persistently oversensitive pixels. And on looking at daylight shots, these pixels are not visible in magnified crops. Only those which correlate with high confidence are considered matches, so there are perhaps twice as many which are suspicious.
+ =