In this class, we continued the topic of noise in CCD images with a discussion of an image of a light source, specifically the flat lamp. The point was to demonstrate by example that the photon noise is proportional to the square root of the number of photons counted.

Here is one of the flat images we used: 20050828.006.fits.

Here is a plot of column 432 of this image (white trace) overplotted with the flat image that was taken next (20050828.007.fits) in red. The two traces are similar but not identical. You can see random differences between the two (noise) as well as a general trend in both (instrument sensitivity as a function of wavelength) and localized features (CCD defects).

When we divide #7 by #6, essentially all that remains is the random fluctuations; the non-random features are multiplicative factors that cancel out.

Then we repeated this process with a different row of the same images where the signal is lower.

In the quotient spectrum, you can see that the fluctuations are now larger in relation to the quotient; the signal-to-noise ratio is lower.
In both quotient images, we used the 'm' key in the IRAF task splot to determine the signal-to-noise ratio in the quotient images. We also used the same command to determine the mean signal in the original images. Then we found that the signal-to-noise ratio was in proportion to the square root of the original signal - as expected for photon counts.