Return to Previous Section
The error bars on the data in Figs. 10 and 11 are relatively large. These error bars do not reflect the uncertainty of the mean (which is typically vanishingly small given the sample size of ~1.6 million pixels). They reflect the dispersion about the mean. And it is a bit disingenuous to treat this dispersion as a statistical error. What it really reflects is the systematic pixel-to-pixel sensitivity variation. In other words, one cannot generally apply a "dark correction" simply by evaluating a mean dark level, and subtracting that constant from an image. If dark correction is needed, one must create the best 2-D dark image possible, and subtract that from one's data.
Dark frames typically contain two sorts of structure. One is typically called "hot pixels" -- individual isolated pixels that have dark currents that are much higher than the typical value. The other is some sort of large-scale fixed pattern. Our detector shows both of these types of structure, as revealed in Figure 12.
The image is a reverse grey-scale. So the hot pixels show up as black specks. But not all the small black specks are hot pixels. Some are due to cosmic rays that struck the detector during the integration. A particularly good example of a cosmic-ray hit is near the top center of the figure. If one integrates a CCD for any significant time, one will accumulate cosmic-ray hits in the image. I shall discuss how to correct for this problem in the next section.
In addition to the hot pixels, there is a global gradient: The bottom of the detector has a higher dark current than the top. And there is intermediate-scale horizontal banding. As an example of how correctable these features are, I show the ratio of two dark frames in Figure 13. Note that the dark-current features on all three scales (hot pixels, intermediate-scale banding, and global gradient) are all quite repeatable from night to night. The only residual features are those due to the cosmic ray hits in the two dark frames. This means that we should not generally have to worry about variations in the dark frame structure within a given night.
The importance of taking darks at the same temperature as the sky frames one wishes to correct is demonstrated in Figure 14. Here I show the ratio of two darks taken on the same night, but at different temperatures. The hot pixels and intermediate-scale banding are not apparent, but there is clearly a residual global gradient. This is in addition to the variation of the mean dark rate discussed in section 3.3 above.
Procede to Next Section
Return to Main Dark Page
Return to Outline