7.5 A Road Map for Optimizing Observations

Dithering and CR-SPLITing more than the minimum recommended values tends to yield higher quality images with fewer residual detector defects, hot pixels, or CR signatures in the final combined image. Dithering is recommended over CR-SPLITs since it allows the removal of both detector artifacts (hot pixels, bad columns, etc.) and cosmic rays. Unfortunately, splitting a given exposure time into several exposures reduces its signal-to-noise when the sub-exposures are readnoise limited.

Broadband and grism WFC exposures longer than about 300 seconds are usually background limited (backgrounds >20e), while medium- and narrow-band images are readnoise limited for all practical exposure times. Thus, the optimal number of sub-exposures is a result of a trade-off between completeness of the unstable hot pixel elimination, CR-rejection, final image quality, and optimal S/N. A schematic flow chart of this trade-off is given in Figure 7.2. The main steps in this, possibly iterative, process are the following:

  1. Determine the exposure time required to achieve the desired S/N.
  2. Determine the maximum number of acceptable residual CRs in the final combined image. This number depends critically on the scientific objective. For example, for a survey of distant galaxies or a globular cluster color magnitude diagram, a few residual CRs will not compromise the scientific output of the observations. In contrast, for a search for an optical counterpart of a radio- or gamma ray-selected object, even one residual CR would not be acceptable over the region of interest. In this latter case, since we expect about ~4% to 7% of the pixels to be affected by CR hits during a one-orbit exposure on the WFC, the requirement that no pixel in the final drizzle stack be affected by CR hits would force one to use at least 4 sub-exposures. For an experiment in which the number of allowed false alarms is zero (e.g., a search for cosmological supernovae), observers may wish to further increase the number of sub-exposures.
  3. Note that even a few thousand residual CR hits cover but a tiny fraction of the 16 megapixel area of the full-frame WFC. In general, the number of pixels affected by coincident CR hits for a given total exposure time and number of sub-exposures N will be:

    (1) \left( 0.05 \times \frac{\mathrm{Exposure\ Time}}{2400\mathrm{s} \times N} \right)^N \times 4096^2
  4. Determine whether dithering is required. CR-SPLITs do not mitigate unstable hot pixels, sink pixels, or other detector defects. If such features would critically affect the science, then dithering is required to remove them. Hot pixels currently cover nearly 2% of the WFC CCDs though <10% of them are unstable (see Section 4.3.5). For some imaging programs, the spatial resolution provided by the WFC and the presence of some detector defects and hot pixels in the final image are acceptable. For such observations, dithering would not be required and one would simply split the exposure time for CR correction. For observations where several orbits worth of data are obtained with each filter, the best strategy is to observe using a sub-pixel dither pattern without obtaining multiple images at each position. Due to the increase of detector defects of the WFC CCDs with age, we only recommend the use of CR-SPLITs over dithering when it is absolutely essential that the sub-exposures place the source of interest at exactly the same location on the detector.

  5. Once the required number of individual exposures has been established on the basis of CR rejection and dithering requirements, the observer will need to verify whether the resulting readout noise affects the achieved S/N.


Figure 7.2: Schematic flowchart of the CR-SPLIT vs. dithering vs. S/N trade-off