Fig. 9-18. Declination Versus Frame Number as Determined From Data in Fig. 9-17 Using Nominal Sensor Parameters. (Slope of solutions in central region is caused by biases in the sensor parameters.)
ends of the data sample are eliminated and the best available-solution for the central portion of the data is obtained. Predicted-versus-observed data plots are then used to refine the selection process in an iterative manner and other plots are used as needed to check the consistency of the results. Thé need for an accurate attitude solution in step 2 of the process is shown in Fig. 9-17(b),* which compares the observed data with predicted data based on nominal sensor parameters and the data from the central portion of the pass. Particularly on the right side of the figure, it is clear that the anomaly involves both Earth-in and -out and that it extends at least somewhat beyond the end of the "pagoda." However, it is impossible to precisely determine the invalid data from Fig. 9-17(b) because of the poor overall fit, even in the central region. (A solution based on all the data yields an even worse fit.)
Figures 9-17(c) (showing the fit to the central portion of the data) and 9-17(d) (showing the pagoda characteristics) compare the observed data with the predicted data based on results from the central portion of the data pass using attitude and sensor bias parameters obtained from a bias determination subsystem similar to that described in Section 21.2. Once an accurate fit to the data has been obtained, the general character of the pagoda effect becomes clear. Both Earth-in and -out begin varying systematically from predicted values when the Earth width, or the difference between Earth-in and -out, drops below about 20 deg. At an Earth width of 12 deg, the Earth-out data turn sharply upward.* (The small ripple most noticeable in the Earth-out data on the left of Fig. 9-17(d) is not a plotting artifact; although the cause is unknown, it may result from variations in the height of the Earth's atmosphere in the infrared.)
'The attitude solution is based on data from the central region only. Using this attitude, data are predicted for the full data pass (including the end regions) to provide a visual comparison in order to identify the data anomaly. This procedure was used for Figs. 9-l7(b) through 9-l7(d). ^ The differences quoted and those shown in the figures are in terms of rotation angle. The arc-length separation between Earth-in and -out is about 14 deg when the effect begins and 7.5 deg when the upturn occurs.
At the time of the above analysis, the cause of the pagoda effect was unknown. Subsequent investigation indicated that it is probably due to delays inherent in the sensor electronics, as described in Section 7.4. The results of that analysis, shown in Fig. 7-21, indicate that the data selection described above is at least approximately correct.
The value of predicted-versus-observed data plots as part of the data validation procedure is shown in Fig. 9-19, which illustrates a data pass where the sensor scan cone does not drop off the Earth before reversing direction and moving toward the Earth's center. In Fig. 9-19(a), there is no visually detectable anomaly, although the previous example suggests that there might be a problem at small Earth widths. This is confirmed by the predicted-versus-observed display of Fig. 9-19(b), which reveals the pagoda effect data which must be eliminated to obtain accurate attitude solutions.
W PAGODA IMfCT DATA WHQ4 19 NOT VtSIALLV DfTECTASLE (D) rJMBarriD V1KSU3 OSSINVID MTA. AS FO* FAAT Ul. SMOKMQ PAGOCA
Was this article helpful?