Results from analysis of the ACA acquisition database suggest that it may be possible to refine the star selection algorithm currently in use by SAUSAGE by deemphasizing search spoilers as a selection criteria. This change would push the algorithm towards selecting slightly spoiled stars over faint stars that have a lower overall success rate.
In the current SAUSAGE algorithm a multiplier is applied to the one-sigma AGASC 1.5 magnitude error for both the candidate acquisition star and potential spoiler. If the spoiler is then potentially brighter than the candidate, SAUSAGE rejects the star. The value of the multiplier depends on the stage of the acquisition process. In the first stage SAUSAGE is most stringent with the multiplier set to 3-sigma. In the second stage the multiplier is reduced to 1-sigma. All star magnitudes are "perfectly known" in stage three (0-sigma).
The acquisition database currently has records of acquisitions for several hundred stars with potential spoilers as flagged by starcheck. This database contains information on the actual acquisition process including observed magnitude, position, and whether the star was successfully identified by the OBC.
The scatter above represents the distribution of identified and unidentified stars as a function of magnitude and radial separations. The distribution of unidentified stars is fairly uniform in magnitude space, while the density of identified stars increases as magnitude separation decreases.
Below is a trend plot of star acquisitions as a function of spoiler magnitude separations. Mean one- and three-sigma values are given by taking the average magnitude error of the sample and applying a multiple of two, and six (two stars) respectively. The red lines represent these cutoffs. Error bars represent the Poisson error.
|Multiplier||# inside||# Unidentified||% Unidentified|
|Zero-Sigma||625||36||5.76 +/- .96|
|One-Sigma||497||26||5.23 +/- 1.03|
|Three-Sigma||102||7||6.86 +/- 2.59|
The three-sigma numbers should be treated with care due to small number statistics. Remember that this analysis includes only stars flagged as having a potential spoiler. Outside of three-sigma the percentage of unidentified stars is actually around 2.2%
The data above show that moving the first stage cutoff from three-sigma to one-sigma would result in a doubling of unidentified stars due to search spoilers. However, dropping the second stage cutoff to zero-sigma would have a negligible impact, and allow the star selection algorithm to choose higher success rate stars overall.
Questions, Comments, Recipes
Email: Brett Unks