IRAS Explanatory Supplement
V. Data Reduction
D. Point Source Confirmation
D.4 Known Source Correlation
Table of Contents | Index | Previous Section | Next Section
- Known Source Prediction
- Correlation of Observations With Predictions
- Known Source Flux and Position Discrepancy Analysis
D.4.a Known Source Prediction
The a priori fluxes in the IRAS bands for the IRC, AFGL, and K stars were estimated by blackbody curves run through measured flux points. The K star estimates were adjusted by global rescaling after a month's accumulation of data. The fluxes for the solar system objects were obtained via standard thermal models. The low-resolution spectrometer objects had no flux estimates.
For each survey scan, the known objects which were to be covered were predicted. The predictions involved the detectors which would be crossed by each source image, the times of these crossings, the flux that should be observed, and a measure of the probability that a detection would result. The scan parameters and the known positions of the sources were used with the detailed boresight pointing history to calculate the geometrical predictions, and once the detectors were identified, their sensitivities and the intersections of their slots with the image blur were used to obtain the expected flux values. The probability of detection was based on the expected signal-to-noise ratio and photometric error, along with the uncertainty in the reconstructed solar aspect angle, the clearance in the slot of the image center, and the width of the image blur. The probability increased with higher signal-to-noise ratio, as photometric error became less likely to thwart the detection process, and it decreased with smaller slot clearance, as cross-scan limit cycling and image blur width became more likely to prevent the necessary amount of flux from arriving at the detector. For the solar system objects, orbital position calculations were necessary. These were computed in heliocentric ecliptic coordinates at the time corresponding to the middle of the observation period and transformed to spacecraft-centered position angles. The remainder of the task was the same as for inertially fixed sources.
After each anticipated known source image had been mapped through the focal plane, the number of detections predicted for it was checked to determine whether the source could reasonably be expected to be detected. The flux predicted for each detector was required to be above a certain level corresponding to the detection threshold on that detector. At least one SC or NSCF prediction was required before any prediction at all was generated. If any prediction was issued, then the estimated fluxes in all bands, even those below threshold, were passed on to the correlation analysis, although no detectors were associated with any band for which the fluxes were expected to be below the detection threshold.
D.4.b Correlation of Observations With Predictions
Association of an observed source with a predicted known source was done strictly on the basis of position agreement using the non-Gaussian statistics described at the beginning of Section V.D.1 (Fowler and Rolfe 1982). After an identification was made, a subsequent identification of the same prediction with another observation could still occur and have a chance to replace the earlier association if appropriate. In such a case, the two associations were compared and only the better match was kept. Similarly, if one observed source was found to pass the tests for association with more than one prediction, only the best match was kept.
If a predicted source was never matched to an observation, then the predicted probabilities of detection were checked to see whether any were above 0.99 and if so, a warning message issued. A given prediction was considered unable to acquire further matches if the latest observation processed was more than 41' past it in the scan direction, or if another prediction at least five positions further downstream was matched.
D.4.c Known Source Flux and Position Discrepancy Analysis
When final match decisions had been made for each predicted known source, those which were identified with observations were investigated for discrepancies in the flux and position. The various types of known sources mentioned above were all kept separated in the statistical analysis of the discrepancies.
In-scan position discrepancies were used to compute the mean, variance, and statistical significance of this error. The information was used in the focal plane geometrical calibration. Only very small mean errors were found and the variances indicated that the a priori errors were slightly conservative. In addition, data were grouped in cross-scan cells in order to determine whether the mean in-scan error was a function of the cross-scan location of the detectors involved. This would reveal any significant rotational misalignment of the focal plane about the optical axis; an upper limit of less than half a second of arc was found for the impact of this effect on source position reconstruction.
Cross-scan position discrepancies were studied in three groups: sources observed only in one band, sources observed in more than one band and sources containing any triple detections. These groups have distinctive ranges of cross-scan errors, and are listed above in order of decreasing uncertainty. For each group, the mean and variance of the cross-scan position discrepancies were computed. General agreement with a priori values was found.
Table of Contents | Index | Previous Section | Next Section