4.6.4 Dithering

Finally, we wish to implement a dither scheme. At each of the positions defined by the single position, map, or cluster in the target specification section, the telescope is pointed to a series of offset positions that define the dither pattern. The purpose of the dither is to minimize the effect of pixel-to-pixel gain variations, help with cosmic ray identification and removal, and aid in removal of image artifacts on the detector. Several pre-packaged dithering options are available in either the full array or subarray readout modes. Clicking on dither pattern ``yes'' brings up the dither selection dialog illustrated in Fig. 4.13.

Figure 4.13: The IRAC Mapping dither pattern dialog.

The dither patterns for IRAC in full array mode currently include a five-position Gaussian pattern, two sizes of Reuleaux triangle, a 9-point random, a 16-point spiral, and a cycling dither pattern; see the IRAC chapter in the Spitzer Observer's Manual for more information.

For the Gaussian, Random, and Reuleaux patterns, the same dither pattern is performed at each map grid position. The ``cycling'' pattern consists of a table of 311 unique offsets, drawn at random from a Gaussian-normal distribution; for each dither, the observations are obtained at the next entry position in the table. The dither pattern can vary quite a bit depending on the number you choose as the ``starting point in the dither table,'' so if you don't like any given pattern, try changing that number and re-visualize the observation.

For the Cookbook example we are developing here, in the dither selection dialog window (Fig. 4.13), selecting ``Cycling,'' with ``Number of Positions'' set to ``49,'' and ``Starting point in dither table'' set to, for example, 10 would mean that at the first map position, exposures are taken with dither-offsets 10-58. At the second map position, exposures are taken with dither-offsets 59-107 in the dither table.

In addition, the selected dither patterns may be scaled using one of the ``large, medium, small'' options, which select pre-determined scalings of the dither patterns, which for cycling dithers are 5:2:1. The offsets in the Gaussian-normal distribution that generates the cycling pattern can be as large as $\simeq 160$ pixels from the (0.0) position. In general, well-dithered observations will mitigate much of the effect of latents, scattered light, and image artifacts, and so we select the ``medium'' scaling option.