SUSI2

Fringing

SUSI2 Fringing in I#814

Images obtained with the I#814 filter show a strong fringing pattern caused by night sky line interference in the thin CCDs. Since flat fields are obtained with white light, they cannot correct for these fringes. Moreover, the fringes are an additive pattern and should therefore be subtracted, not divided. To remove the fringes, the fringe pattern must be extracted from the scientific images themselves.

Fringes appear quite well in Fourier space. Therefore, checking the power spectrum of the raw and cleaned images is a good way to monitor the cleaning and to adjust the normalization factor used below. Note however, this does not work very well for crowded fields, as the objects contaminate the power spectrum.

The following procedure is will take care of fringing in a simple manner and give good results:

Obtain the observations with a large dithering pattern. Make sure you have at least 5 long exposures in I, ideally at least 5 exposures per field.
Subtract the bias from all frames
Normalize the sky level of all your I exposures to the average sky level (you may want to work field by field). Store the normalization factor for further use
Median average all the I exposures (again, you may want to work field by field). The result should contain no objects, just the fringes and the flatfield structure
Divide the result by the twilight I flat. You are left with a flat sky and the fringes (which have also been divided by the flatfield - we will correct this shortly)
Get the average sky level of this image and subtract it (you want to have an image with a mean ~zero). You are left with the flatfielded fringes alone
Multiply the result by the flatfield: this is your master fringe template
For each image, multiply the master fringe template by the normalization factor you used for that image
Subtract the result from the original image
Divide the defringed frames by the twilight I flat

Dirty image in I Clean image in I

The fringing pattern

You can retrieve the fits file corresponding to a "classical" fringing pattern by clicking on the above image. It was built with 34 I#814 images. The exposure time for all the images was 600s. If you don't want to build your own fringing pattern (we recommend that you do, but if you do not have sufficient data) you can use this one. However, note that you should:

flat field it with your sky flat (the pattern_raw.fits is not flatfielded)
adjust for the exposure time
adjust for the difference in the intensity level:

Take the ratio of the mean level of the science image and the mean level of the fringing pattern as a first approximation
Multiply the fringing pattern by this factor
Subtract this corrected fringing pattern to your images

This was how the above clean image was made. Remember, if you want to do precise photometry, you have to fine tune and adjust the factor for each image.

Here, you find a new fringing pattern. It was obtained combining 5 images (I#814) with exposure time of 1200 sec each, taken on November 10, 2004. It was corrected for the bias but was not flatfielded.The average has been subtracted, so the pattern is varying around zero. We compared the image to the old one, and although the intenisty of the fringes change, we could not find any changes in the shape of the pattern.

Send comments to : ls-susi
Last modified: Thu Nov 18 2004 (Silvia Ragaini)