IRAC: Best Observing Practices

1. Always check your planned observation by overlaying it on an appropriate astronomical survey image such as one from the 2MASS/DSS/MSX surveys. Check for the range of allowed observing dates (visibility windows). Verify that bright objects, e.g., from the 2MASS or IRAS catalogs, will not ruin the planned observations in terms of scattered light, persistent images, image artifacts, or saturation.
2. Use dithering instead of in-place repeats (see section below).
3. Use medium- or large-scale dither patterns unless performing a map with very tight requirements for uniform coverage (see section below).
4. If the field you are planning to observe includes very bright sources, consider using the HDR mode to allow for flux measurements of bright sources and identification of saturating sources (see section below). If you plan to observe a single very bright point source, consider the subarray mode (see section below).
5. Inspect the field being observed, including the "serendipitous" field covered by one pair of detectors while another is observing the target of interest. Residual images can persist for some time and ruin your observation (see section below).
6. If rejecting asteroids is important, plan at least two observations of your target, separated by a few hours (see section below).
7. For the most sensitive observation, use the 100 second frame time. Note that the 200 second frame time is no longer supported; routine sky darks are not taken with 200 second frames.
8. For highest quality images, put your source in the inner 1/3rd of the array.

Dithering vs. in-place repeats

Scattered light from bright sources near the edge of the field of view sometimes cannot be avoided, but it can be prevented from contaminating multiple frames by using a dither pattern larger than the characteristic size of the regions that produce stray light. Scattered light is well-rejected by the post-BCD pipeline if dithering and mapping offsets are large enough, for example, with the medium and large dither patterns (but not small), to ensure that stars in subsequent images are moved out of the scattering zones, and the redundancy is high enough to allow for effective outlier rejection (five dithers is sufficient).

We discourage using in-place repeats (successive frames taken at the same position), especially in observations that have less than ten different dither positions, both for the reasons mentioned above and also because the residual first-frame effect will lead to a bias pattern difference between the "repeats" and the "first frames" of a repeat set. Taking in-place repeats is recommended only in the case of time series measurements with stringent requirements for stability, such as observations of planet transits. In general, the better handling of pixel-to-pixel variations using dithers will more than compensate for the reduced amount of integration time; that is, the realized signal-to-noise level will be higher using N-1 dithered observations than N observations with repeats. (In-place repeats are specified in the AOT window under "For Each Pointing/ Number of Frames;" dithers are specified lower in the AOT window under "Dither Pattern.")

Depending on the number of bright sources in your planned field, and the placement of your observation in the schedule relative to other observations that may have bright sources, there may be persistent images in the arrays during your observations. The calibration data (flats and darks) are carefully planned to avoid such artifacts, and they are erased by anneals every 12 hr. If your observation is contaminated by persistent images, the effect on the final data quality is reduced in a dithered observation by at least a factor of 1/N, where N is the number of dithers. Greater reduction occurs when robust averaging (outlier rejection) is used.

Taking into account possible residuals of the first frame effect and image artifacts due to bright sources, we recommend that sensitive, background-limited maps be made using either as many dithers as possible, or a fine map grid combined with the cycling dither pattern. Shallow, read-noise-limited observations, or raster maps specially designed to identify phenomena with a certain time-dependence (such as asteroid motion), may be best done with combinations of small dither patterns and small map steps. The small scale dither patterns should only be used in conjunction with mapping using 1/3-2/3 array offsets. The small scale dither patterns do not provide sufficient redundancy under other circumstances.

HDR mode

Stellar photometry mode

Subarray

Mapping

Observing near bright targets

A very common and challenging type of observation is to search for a faint source near a much brighter one. The choice of observing strategy and assessment of the technical feasibility for such a search requires great care. First, it is important to determine the brightness of the point response function (PRF) of the main source at the distance where the putative companion is being sought. The Table below gives the intensity of the PRF, normalized to the peak for a source centered in a pixel, at several distances from the center. A well-designed experiment will be able to "cancel" the PRF down to a certain level. For example, if the PRF is down by a factor of one thousand from the peak at the distance being searched, it is plausible to search for features at a level of 1x10^-4 if some type of PRF removal is done. And, furthermore, having knowledge of the potential shape of the companion (e.g., a point source or disk) will obviously discriminate it from the remaining PRF of the bright source. Remember that the Poisson noise of the bright source at the location of the bright target will always limit the accuracy of the PRF removal.

Table: Values are derived from 5x (0.24"/pixel) oversampled PRFs created from observations of calibration stars. I/Io is calculated as the flux density /pixel in a 1 pixel annulus at the respective radii divided by the flux density measured in the central pixel. Encircled Energy is the ratio of the flux density enclosed within the representative apertures to the flux density in a 12.2" aperture.

Here is a recommended strategy for this type of observation. Center the bright source on the array, and observe it with a dither pattern with medium scale factor. Use a pattern with good subpixel sampling, to enable future super-resolution techniques; for example, consider the 9-point Random (1/3-pixel sampling) or the 16-point spiral (1/4-pixel sampling). Use the HDR mode, so that there are short frames at the exact same position as the long frames. These may help with centroiding, because they will be less saturated in the core. Use a long enough frame time for the long frames so that you will gain the required sensitivity on the putative faint source in one (or a few) frames, preferably without saturating the primary source over more than a few pixels. Finally, and probably most importantly, design the experiment as a pair of AORs, with timing constraints such that the source will be seen at two very different roll angles. The range of possible roll angles depends on the ecliptic latitude of the source and must be verified using Spot visualizations near the beginning and end of the visibility windows for a given source. Typically the range of roll angles is about 20 degrees. The PRF will remain fixed in array coordinates, so the basic calibrated images at the two epochs can be subtracted. While any detector artifacts and most of the light from the bright source should subtract out, two images of any real faint object(s) should remain.

Observing single faint sources

Confusion and other background issues

To separate faint asteroids from more distant targets, it is a good idea to observe your field of view at least twice, separated by a few hours at least, if your source lies at a low ecliptic latitude (within 15 degrees of the Ecliptic or so).