[Top] [Prev] [Next] [Bottom]
7.4 Flatfield Residuals
There are currently four UNI (flatfield) files for the f/96 camera at 1360, 4800, 5600, and 6600 Å and two UNI files for the f/48 camera at 3345 and 4800 Å. The UNI files have been derived from heavily smoothed flatfields. Thus, they do not flatten small-scale features, such as scratches and reseau marks, that exist in the flatfield response and can affect your photometric accuracy.
How much the small scale features affect the accuracy depends greatly on the type of data and the method of analysis. In some cases, careful treatment can improve the calibration. Figures 7.2 and 7.3 show relatively high signal-to-noise full-format flatfields obtained in the UV for the f/96 and f/48 cameras, respectively. Many of the features to be discussed here are evident in those figures.
7.4.1 Border Effects
The borders of FOC images suffer from corruptions arising both inside and outside the detectors. Among the most obvious external effects are the finger-like shadows cast by the occulting fingers (two occulting fingers for f/96 and the slit location finger for f/48.) In addition, square masks in front of both detectors shadow the upper left and lower left corners of the f/96 image (upper and lower left) as well as the lower right corner of the f/48 image. Furthermore, geometric correction transforms the straight edges of the original raw images into curved edges, most noticeable on the left and right sides.
Internal border effects show up in a few bad rows at the top and bottom of the raw image and the left-most columns of the raw image as well as a significant number of columns at the beginning of the scan line (right side of the image). In all FOC images, the internal border effects are present regardless of format; however, they do change from one format to another. In particular, the corrupted pixels at the beginning of the scan line arise from defects in the beginning of the sawtooth in the scanning waveform. The corrupted beginning is about 5% of the scan line for most f/96 formats In the f/48 detector it gets progressively worse for smaller formats (from about 5% for the full format to about 25% for the 128 x 128 format). The horizontal stripes seen in the bottom left of the f/96 image result from a ripple instability of the coil drivers at the beginning of a frame scan. None of these effects are normally correctable.
Figure 7.2: f/96 External UV Flatfield Image
Figure 7.3: f/48 External UV Flatfield Image
7.4.2 Video and Digitizing Defects
The narrow line running from the bottom left corner to the upper right corner (clearly visible for f/48, less so for f/96) is due to the read beam of the television camera not being completely blanked before it flies back to the beginning line at the end of a frame scan. This effect, along with a change in path, becomes more noticeable in smaller formats. The narrow horizontal features at the right edge, especially at lines 256, 512, and 768, are due to noise glitches on the scan coil driver caused by changes in the most significant bits of the line counter. The central 512 x 512 pixels in both cameras are outlined by sharp changes in sensitivity. Heavy use of the 512 x 512 format has burned a charge discontinuity into the camera target array at the edges of this format. None of these effects is normally correctable and the affected areas should be treated as bad pixels.
7.4.3 Reseau Marks, Scratches, and Blemishes
A regular grid of reseau marks used to measure detector distortion spans both detectors' photocathodes. These reseau marks have about 90% opacity and are not normally worth trying to flatfield. In addition to the reseau marks, there are various scratches and blemishes, much more numerous in the f/96 camera. The scratches and blemishes generally appear much deeper in the far-UV-as much as 30% opacity for some scratches. Because the pipeline flatfield correction is heavily smoothed, none of these effects will be flatfielded out. Hence, photometry of sources which fall on or near these image defects can be compromised.
The imedit task in the images package or the rremovex task in focphot package can be used to repair such cosmetic defects in images having a source that falls on a reseau mark or small scale blemish. These tasks replace the values of the affected pixels with the average values of their neighboring pixels. Great care, however, must be taken in interpreting photometric results for sources which are directly affected by such image defects (i.e., in which the peak of the source falls on or immediately adjacent to an image defect).
7.4.4 Pattern Noise
Pattern noise, neither fixed nor constant in magnitude, constitutes another source of non-uniformity. Two types of patterns are often present, although not always easily seen in low count extended areas or flatfields. The more noticeable one is an approximately sinusoidal pattern with its peaks and troughs oriented at an approximately 45 degree angle and a period of 3.35 pixels for f/96 (it is just barely discernible in Figure 7.2). It is believed to originate from a moiré effect between a TV tube grid and the diode array on the target. The amplitude of the pattern depends on the count rate in the area. In flatfields with count rates between about 0.02 and 0.1 counts pixel-1 s-1 for a 512 x 512 format, the rms amplitude of the pattern is about 5% of the flatfield counts for f/96 and about 2.5% for f/48 (the peak deviations from a flat response due to this pattern are at least twice these values). At lower count rates, threshold unknown at this time, the pattern disappears. On the other hand, the pattern intensifies when count rates are in the nonlinear regime and thus is much more easily seen. In fact, it is a quick way of recognizing serious nonlinearity in an image.
A second pattern arises from some form of interference with an FOC digital timing waveform that has a four-pixel period. It shows up as vertically striped patterns on the flatfields (visible in Figure 7.2). Although very coherent in orientation and frequency (in the raw image), the details of the modulation do not appear to remain constant in either phase, waveshape, or amplitude from image to image. The rms amplitude of this pattern in moderate count-rate flatfields, is approximately 2.5% for both cameras. Like the 45 degree pattern, this pattern seems to disappear at low count rates.
Given the nonlinear nature of the amplitude of these patterns and their variability in position (phase), there is no general method for correcting them. When count rates are moderate across most of the image, i.e., from an extended object or PSF halos, Fourier techniques can sometimes proves useful in removing the pattern. The main purpose of these techniques should be viewed as providing aesthetically pleasing images rather than as improving photometric accuracy.
7.4.5 Large Scale Variations
Large scale variations are those spatial variations having relatively low spatial frequencies, i.e., 20 or more pixels. The UNICORR step in the pipeline attempts to remove such variations from the image. Large scale variations in the response of the FOC do not appear to depend strongly on wavelength between 1300 and 6000 Å; generally speaking, the large scale response does not change more than 10% for all pixels except at the edges and corner of the full format. Beyond 6000 Å, the flatfields begin to change significantly, generally with poorer relative sensitivity towards the corners.
Obtaining flatfields in the UV requires a great deal of spacecraft time for each wavelength desired. At the moment, only one UV flatfield each exists for the f/96 and f/48 camera (at 1360 and 3727 Å respectively). It is not likely that there will be any more UV flatfields obtained for f/48.
The f/96 large scale response appears to be accurate to 1 to 2% rms over the most of the photocathode at the wavelength where it was obtained, excluding the edges and corners, and regions where the scanning oscillations are significant. The accuracy for f/48 is estimated to be 2 to 4% rms over comparable areas.
7.4.6 Time Variability
A small amount of temporal variability has been observed in the flatfield response; it is largest just after the FOC is turned on and begins taking exposures. Changes of about 1 to 2% are seen with respect to the flatfield response after an hour of exposures. The changes for f/48 are about twice as large. In general the response at turn on is higher at the center and weaker at the edges of the full format.
7.4.7 Format-Dependent Effects
The FOC flatfield depends on the video format used (Greenfield and Giaretta, 1987, FOC ISR 024). You cannot just divide an image by a flatfield derived from the corresponding subsection of the full-format field, even if you take great care to align the two images so that the reseau marks overlap. This effect was suspected to be due in part to the limited resolution of the geometric distortion field provided by the reseau marks and the resulting change in the apparent pixel size with position. More detailed analysis by Greenfield using the new geometric correction method described on page 6-5 showed that these suspicions were ungrounded. The variations in sensitivity with position truly depend on the video format. At this time, however, the appropriate correction files have not been derived, although the possibility of applying a format-dependent flatfield does exist within the current FOC pipeline.
[Top] [Prev] [Next] [Bottom]
stevens@stsci.edu
Copyright © 1997, Association of Universities for Research in Astronomy. All rights
reserved.
Last updated: 11/13/97 16:44:29