46.6 Flatfield Anomalies
The photometric accuracy of the data is often limited by the quality of the flatfield used to calibrate the data (see "Determining the "Best" Reference Files" on page 45-14 as well as the WWW Reference File and Closure Flat Memos). Details of the flatfields and delta flats available for calibration were discussed in "Determining the "Best" Reference Files" on page 45-14. In this section, we summarize residual features that may remain; we recommend experimenting with different flatfields to determine which are best for the data (see "Improving the Flatfield Correction" on page 46-15).
- Changes due to decontaminations. The flatfields are relatively stable between decontamination procedures; however, they do undergo some incremental changes across decontaminations. These changes are usually at the few percent level and so, may not be important for some observations1; if necessary, the changes may be corrected using an appropriate delta flat (see "Choosing and Generating Delta Flats" on page 45-18).
- Time variability due to scattered light changes. Following a decontamination, the gradual buildup of contaminants causes an increase in the scattered light (see "Scattered Light" on page 46-2); the corners of the internal flatfields used to monitor the instrument performance are seen to droop relative to the chip centers. Note also that the dust features change character in response to the increase in scattered light.
- UV throughput decreases over time. The buildup of contaminants responsible for the increase in scattered light also causes a corresponding decrease in throughput (see "Photometry" on page 46-9 and WF/PC-1 Instrument Science Report 93-02).
- Persistent measles after February 1992. These features are at the ~1% rms level, although the cores can deviate by up to 5% or more. See "Persistent Measles" on page 46-2 for more details.
- Large gradients caused by F122M. Due to the Earth's albedo, some broadband filters had to be taken in conjunction with a neutral density filter (either F8ND or F122M) to obtain sufficient numbers of unsaturated earthflat observations for creating reference files. The F122M filter, designed as a short wavelength pass filter, had a red leak that was used to provide neutral density at long wavelengths. The filter consisted of a very weak MgF lens, with a dielectric short pass filter deposited on the side closest to the pick-off mirror. However, the dielectric coating was non-uniform, resulting in a 25-30% brightness gradient running across the WF field, which is very weakly dependent on wavelength within the visual range. There were also several small pinholes in the F122M filter, which caused wavelength dependent "doughnuts" (images of camera relay optics pupils) to appear in the flats with strengths of 2% to 5%.
- Anomalies due to the F8ND filter. In other flatfields, the F8ND filter was used as the neutral density filter; it consisted of a strongly curved, thin meniscus whose center of curvature was at the pyramid apex. The surface was coated with a highly reflective metal film, which provided a neutral density uniform to less than 1% in intensity. The problem was that strong reflections could occur when F8ND was crossed with certain other filters. The strongly curved surface, together with its high reflectivity, caused light reflected from other filters to be strongly concentrated in a circular pattern centered near the pyramid apex. These reflections were formed by light passing through F8ND, reflecting off the other filter, reflecting from the curved surface of F8ND, and then passing through the other filter and onto the CCD. The strength of the circular reflection depended on the filter composition, and varied from <1% to ~30%. These reflections caused the F8ND filter to be dismissed as unusable early in the mission, though more recent tests show it is capable of giving near-perfect flats in certain filters. For filters which consist only of anti-reflection coated colored glass (polarizers, F555W, F569W, F725LP, F785LP, and F850LP), the intensity of the filter reflection is less than 1%. Certain other filter constructions gave very weak reflections, such as F336W, where the UG-11 colored glass component had a low throughput and was between F8ND and the F336W thin film component; in this case, the filter reflection is reduced to 1-2% in intensity (depending on field position).
Note that the neutral density filter effects have not been removed from the SV and the non-SV epoch flatfields; to avoid the problem, use a flatfield close in wavelength that was not generated with the F122M filter or use a -closure or a high fidelity flat2. For further suggestions on improving the flatfields, see "Data Accuracies and Problem Solving" on page 46-13 (see also OV/SV Report, Faber et al., 1992, Chapter 6, and Closure Flatfield Memo on WWW). - Individual filter anomalies - short exposure reciprocity effects. For exposures shorter than 1 second, the sensitivity of the CCD corners tends to increase. The effect is largest in the blue (~5%), and diminishes towards red wavelengths. The effect is strongest for 0.11 second exposures, and is nearly gone for exposure times of over 1 second. Hence it is useful to optimize the exposure times of the flats to match the data being flat-fielded; the Closure Flatfield Memo lists flats optimized to short and longer exposures.
1WF/PC-1 ISR 92-4.
2 Note that the use of a different flat will affect the photometric calibration.
Copyright © 1997, Association of Universities for Research in Astronomy. All rights reserved.