IRAS Explanatory Supplement
V. Data Reduction
D. Point Source Confirmation
D.2 Overview of Seconds-Confirmation
D.2.i In-Band Seconds-Confirmation Photometric Refinement
Table of Contents | Index | Previous Section | Next Section
Photometric refinement was performed for pairs of detections which passed several qualification rules. When performed Gaussian estimation was used, where the Gaussian error was assumed to be in the logarithm of the flux. The estimates for the a priori errors were derived from analysis of a posteriori discrepancy dispersions as functions of signal-to-noise ratio in each band and the combination of the Gaussian assumption with well-modeled variances was found to be quite good in subsequent
2.
To be eligible for flux refinement processing, the ratio of the fainter detection to the brighter one had to be greater than 0.67; otherwise the brighter detection alone was used. In cases of confusion (except detections discarded in a previously processed case), only candidate fluxes were used because confusion was detected only when it was the drop-dead detection that was composed of more than one source image. When the detections were unconfused and met the flux ratio requirement, they were still required not to show saturation of the analog-to-digital converter if both detections were saturated then the brighter flux was used. If one and only one was saturated then the unsaturated detection alone was used. Finally, if the detections were on edge-overlap slots opposite a dead redundant detector (or a significantly noisy one which produced no detection conformable with the pair being processed), then only the brighter detection was used for flux information, unless it was saturated in which case the fainter detection alone was used.
The drop-dead detection always provided the source name (time tag and detector number). The numbers of all detectors involved were always recorded in the arrays set up for this purpose (see Table X.B.6). When only one detection was used for flux information, it also supplied the signal-to-noise ratio, detection correlation coefficient, and baseline value. When refinement was done, the maximum signal-to-noise ratio and detection correlation coefficient were retained and the unweighted average of the two baselines was kept.
The order of the detector numbers in the array was determined as follows. In the double-detection mode, the first number is the candidate, and the second one is the drop-dead unless one and only one detection was saturated in which case the first number is that of the saturated detector. There is no way to tell the difference between both detections being saturated and only the candidate being saturated without recourse to additional data. In the triple-detection mode, the detector numbers are in decreasing detection time order if there was no analog-to-digital converter saturation involved. Otherwise the first number is the fainter edge-overlap detector, and the other two follow the rules for double-detection mode number order.
Table of Contents | Index | Previous Section | Next Section