Friday 6 December 2013

On the Fringes

I have been investigating some ways to further improve the automatic post-processing of images from the telescope. While I haven't really succeeded in that goal, the results are still interesting.This has to do with the swirling patterns seen on some images, particularly when imaged close to the horizon. In my previous post on flat-fields, these patterns were visible on the resulting image of M51.

These patterns, as I have found out from some helpful people at Bolton AS, are called fringing. There isn't much information on the net about this as it's only a problem that afflicts the very sensitive back-thinned CCD's, which are generally cost more that interested amateurs are willing to pay.

It occurs as a result of the extreme thinness of the silicon boule overlaying the CCD. This thinness allows very high quantum efficiency to be achieved, but it also means that some longer wavelengths of light (redder wavelengths) are not always absorbed by the silicon. Instead, some light passes straight through, bouncing off the back of the chip, passing through it again and rebounding back from the front of the silicon where it first entered. This light then interferes with the incoming light now entering the chip. 

If the reflecting light has travelled an integer number of wavelengths before meeting the new incoming light, the two are in phase and interfere constructively, adding together to produce a brighter spot.
If the reflecting light travels half a wavelength out from a full wavelength, the two are out of phase and interfere destructively, cancelling out to a degree, resulting in a darker spot.
Offsets between these extremes create a varying degree of construction/destruction. And which one of these you get depends on the precise thickness of the silicon at that point.

Since modern manufacturing methods, while very accurate, cannot guarantee smoothness to within quarters of the wavelength of light (about 100th the thickness of a human hair), you get some areas of the chip where constructive interference occurs and some areas where destructive interference occurs. These show up as the swirling patterns seen on some images, which are like contour lines on a map.

I have tried to isolate this fringing pattern purely using all your images taken with the telescope. This time I used the most recent 2,000 radec images from the telescope, firstly to create a fantastic flat-field by normalizing and then median stacking the images. Once this flat-field was created and smoothed out a little, I applied it to the 2,000 source images. These flatted images were again normalized and median stacked to produce this:

Essentially a contour map of Galaxy Cam's CCD

The fact that it is possible to measure the precise shape of a CCD in Tenerife to within tiny fractions of the wavelength of light, without so much as taking out of its camera housing, purely from minute and usually invisible variations in the images it takes still astonishes me.
You can see the undulations in the chip, as well as the tracks that the machine that polished the chip made during manufacturing.

Unfortunately, returning to more practical matters, it is almost impossible to automatically use this to reliably improve images. Because this pattern is dependent on the wavelength of light causing the interference, the patterns not only varies depending on the filter used, but also what colour the sky happened to be. There are various sources of background light. Moonlight, sunlight (at dawn and dusk), artificial lights, airglow, all of which contribute different wavelength to differing degrees at different times, even within the same night. The pattern above is something of an average. Individual images show shifted patterns, with the fringes slightly out of line. Many images seemed to show some fringes more strongly at one side of the picture than the other.

I have concluded that fixing this problem is, for now, beyond the fringe (pun intended) of what can be automatically accounted for in a real life situation, with a telescope taking as widely varying pictures as ours does.

On the plus side, I hope to use this map to subtract some fringing patterns from the current flat-field, which are in some cases only exacerbating the problem. If you would like the FITS data for this fringe map for your personal use, you can download it here