9.6 Flat Fields

A number of sky and dome flats had been taken, as listed in Table . Additional data, such as exposure time and level, can be found in Table , , and , although the ordering of the images in the JB and JU tables is not chronological, i.e. the dfsc numbers can not be inferred from them.

 

Table: The Available Photometry Flat Fields

Type	Filter	#frames	dfsc numbers	night(s)
Sky flat	Gunn r	14	1362-1364, 1498-1501, 2175-2181	1, 2, 7
Sky flat	Johnson B	14	1365-1366, 1502-1505, 2048-2052, 2182-2184	1, 2, 6, 7
Sky flat	Johnson U	16	1367-1369, 1506-1510, 2053-2056, 2185-2188	1, 2, 6, 7
Sky flat	Johnson V	0
Dome flat	Gunn r	10	1808-1817	5
Dome flat	Johnson B	10	1670-1679	4
Dome flat	Johnson U	20	1649-1668	4
Dome flat	Johnson V	14	2515-2528	11

 

The 44 sky flats were corrected for bias, dark, shutter, and fat zero. The images from before the CCD temperature change (i.e. the flat fields from night 1 and 2) did not get the extra fat zero correction, since it was not known at that time.

A thorough comparison of the flat fields from the different nights showed, that they were not the same, there were gradients in their quotients. While the flats within night 2+ (i.e. night 2, 6, and 7) agreed well (within 0.2% in GR and < 0.1% in JB and JU), the agreement between night 2+ and night 1 was not good, 0.4% in GR and 0.15% in JB and JU. Therefore it was decided to make one set (GR, JB, JU) of flats for night 1, and one set for night 2+.

In each of the 3 filters, the images from night 1 and 2+ were combined separately (weighting by the level in the images, and using a 2 sigma clipping) into a night 1 and 2+ flat, respectively. Care was taken that objects (stars) did not make it to the combined flat.

Since the night 1 flats are made from only 2 or 3 images, the noise in them are too big. Therefore we do the following: we assume, that the pixel-to-pixel variation does not change from night 1 to 2+, but only the large-scale variation. We therefore make the 3 quotients between the night 1 and the night 2+ flats, and fit the large-scale pattern in this using a many piece (6 in x and 10 in y) cubic spline. This fit is then multiplied with the flat from night 2+, thereby getting flats with the pixel-to-pixel variation from night 2+, and the large-scale variation from night 1.