Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.mrao.cam.ac.uk/wp-content/uploads/2014/04/AGrange.pdf Дата изменения: Wed Apr 16 18:57:14 2014 Дата индексирования: Sun Apr 10 13:27:02 2016 Кодировка: Поисковые слова: annular solar eclipse

Characterising Deep Depleted CCDs for the NGTS Project
Andy Grange, Dr Michael Goad
Introduction
The Next Generation Transit Survey is a new wide-angle ground-based survey optimised for finding Neptunes and Super-Earths around nearby bright stars of spectral types K and M via the transit method [1]. NGTS comprises 12 telescopes with 8 sq degrees FOV, F2.8 (Astrograph), on independent mounts (OMI) and new red-sensitive deep-depleted CCDs (Andor). Key to the detection of small transiting planets is high precision (mmag) photometry. Previous surveys such as the highly successful WASP survey have limited photometric precision due to a variety of design and operational issues including temperature dependent focusing effects, and correlated (red)-noise. These effects have been engineered out of NGTS. Key to achieving this goal is the use of custom built telescopes with individual mounts combined with precision autoguiding to keep individual stars on a single pixel. We also take care to characterise the flat fields, especially their colour dependence, to very high precision which enables high precision photometry even without autoguiding. Prototyping of the NGTS instruments at Geneva has shown that the data are white-noise limited. Here we present the results of the lab tests performed at Leicester.

The Blue Diamond Effect
The blue diamond effect is caused by surface imperfections on the silicon chip introduced during manufacture, resulting in a repeated large scale pattern in the flat field at blue wavelengths. As they are a surface property, these effects diminish and eventually disappear completely when observing at red wavelengths (compare figures 2 and 3).

Other Defects
Our tests have also revealed other CCD defects that we can now calibrate out. These include hotspots and small surface defects such as scratches on the surface of the silicon and coating residue. These features are within specification and can be flat fielded out in a colour sensitive manner.

Laboratory Flat Fielding
High precision photometry requires accurate flat fielding. However flats are generally taken at twilight when the sky is blue while our target stars for this survey are red. Thus the blue diamond effect apparent in the blue flats will introduce a systematic error in the light curves. We can remove the blue diamond by using a colour appropriate flat field produced in the lab. Flat fields are produced in the lab using a raster scan, illuminating a small area of the chip at a time. By averaging over your illumination and dividing it out at each position an illumination independent flat field is produced. As we have stable illumination (ELP and LED) we can stack frames over long time periods, achieving over a million counts per pixel. Achieving this means that you can accurately accurately characterise and thereby remove the blue diamond effect (~1%), pixel to pixel variations (~1.6%) and any colour dependent defects. The main challenge of producing an accurate flat field is in the precise registration of the individual scans. A 0.1 pixel misalignment introduces easily visible artefacts.

Figure 3: Division of 450nm by 650nm lab flat fields of different cameras. In the top frame is a mark from cleaning residue. In the bottom left there is a colour dependent surface defect (scratch) and the bottom right panel shows a colour independent CCD defect.

Linearity and Readnoise
Readnoise and dark current are accurately measured in the lab by taking several thousand bias and dark frames. Readnoise levels are around 10 electrons per pixel in expected observing conditions (3MHz readout, 1.5s readout time, gain x4 e-/ADU). The stability of the bias was also examined and found to be highly stable. The stability of all aspects of CCD performance can be monitored whilst on sky with reference to the lab data, tracking bias levels and dark levels, bad pixels and flat fields. Linearity tests are performed to check how accurately observed counts maps to incident light levels. These chips are found to be highly linear (within 1%) up to the full well depth, as shown below. This is important for accurate photometry of very bright and very dim stars.

Figure 1: A sample raster scan, moving a square to illuminate all of the CCD

Figure 4: Plots of Counts against exposure time, using a constant illumination. The linear section is fitted with a straight line, and the residuals shown.

Shutter Correction
As we are targeting bright stars we have short exposure times and the finite shutter time is an important correction, especially at low light levels. It is a small correction to the flux to account for the slight differential exposure time introduced by the finite shutter opening times [2]. Again, the stability of this feature will be monitored on sky.
Figure 2: Division of the lab flat field at 450nm by flat at 650nm, clearly showing the blue diamond effect Figure 5: Shutter correction to be applied to all images

[1] Wheatley et al, Proceedings of the International Astronomy Union / Volume 8 / Symposium S299 / June 2013, pp 311-312 [2] Zissell, 2000Journal of the American Association of Variable Star Observers, Vol. 28, p. 149-156

The NGTS consortium comprises four UK universities (Warwick, Leicester, Belfast, Cambridge) with partner institutes in Germany (DLR, Berlin) and Switzerland (Geneva Observatory)