De-striping and Flux Bias
Baseline removal and cross-scan offset de-striping are discussed in the section on LAUNDR.
Simply applying baseline removal results in a lot of negative data (all the points below the baseline). Since handling negative data properly in HIRES requires some special attention to certain processing parameters, negative data are discarded by default. Therefore, baseline removal can result in large areas of artificially low (or no) coverage. Throwing out so much data can cause a large number of artifacts, reduce the photometric reliability, and always creates a positive bias on the noise. On the other hand, cross-scan offset removal often leaves a high DC background. This can also cause problems: point sources converge slowly and often "ring" on high backgrounds. (See section on HIRES artifacts for more information and examples of ringing and other artifacts.)
For these reasons, the flux bias parameter (FBIAS) may be used to add (for baseline removal) or subtract (for cross-scan offset) a DC value to all the data before HIRES processing and compensate for it again afterwards. The default procedure is to calculate an optimum FBIAS value during the data preparation process that ensures the data are positive but as close to zero as possible during HIRES processing. By default an FBIAS is chosen such that the lowest 1% of the data are discarded.
Note that the use of FBIAS is transparent--- the bias is applied only during processing and is removed from the output maps.
Field Size, Pixel Size and Wavelength
HIRES fields may be any size up to two degrees square. Currently a 1 degree field processed to twenty iterations at all four bands requires roughly two hours or more (depending on the field) of processing time on a SPARC 1. Pixel size is set to a default of 15", although other sizes can be specified. This is several times smaller than the best achievable resolution of HIRES.
Fields are normally processed at all four IRAS wavelengths (12, 25, 60, and 100 µm), but a subset may be specified.
IRAS surveyed most of the sky with three separate hours-confirming observations. The section on the IRAS spacecraft and mission contains an overview of the survey strategy. HCONs 1 and 2 were concurrent, HCON 3 was done several months later and only covers about 2/3 of the sky; because there is generally a larger time difference between HCON 1 (or 2) and HCON 3, there is a greater difference in the zodiacal background. Because of flux-dependent non-linear response variations of the detectors this can lead to a calibration difference for HCON 3 which can be significant.
HIRES can process data from any combination of HCONs. Popular combinations are 1 and 2 together, 3 alone, which can be useful in examining the effects of the background on processing, or 1, 2 and 3 all together. HIRES may occasionally have difficulty converging to a good image if the third HCON is included, especially when cross-scan offset de-striping is used.
Number of Iterations
It is possible to set the number of iterations for HIRES, and also to choose which iteration intervals have maps produced.
By default HIRES is run for 20 iterations, and maps are produced at iterations number 1, 5, 10, and 20. The limit of 20 was chosen because, for the majority of test fields, further iteration tended to show little or no increase in resolution but greater incidence of artifacts.
The correction factor variance (cfv) maps are the best diagnostic for determining if further iteration is warranted. See the section on HIRES output maps for more information on the diagnostic maps. Note that further iteration will appreciably increase the amount of processing time needed, and that output increases alarmingly when maps are produced for each iteration.
Survey vs. Additional Observation (AO) Data
HIRES can process data from any combination of survey data and/or Additional (pointed) Observations. There are advantages and disadvantages to both, and to their combination. Survey data may include multiple scan angles and is more likely to give a more nearly circular beam with accompanying higher resolution. AOs, however, suffer from fewer calibration variations and less striping because each AO field was observed all at once. The raster scan pattern of an AO tends to give a more elliptical beam than for survey data, but the image may converge better and be more reliable. In addition, effective AO coverage is often deeper than survey coverage. In general, it is worth looking at AOs if they are available.
Where they are available, it is possible to stack multiple AOs of a target, or to stack AOs and survey data. In many cases, different AOs had radically different scan angles, which effectively provides much more information, but there may be problems with calibration variation and with artifacting at the edges of the (smaller) AO fields. In practice, most of the benefits are realized after stacking two or three AO s; more slows down the processing dramatically without much improvement. One limitation in using AO data lies in their very small size (0.5 degrees by 1 degree). For more information about using AOs, see the User's Guide to IRAS Pointed Observations or contact IRSA support at IPAC (email@example.com).
AOs suitable for processing are intensity grids observed with the following macros: