Calibration of AGASC 1.6

Introduction

Below is presented a proposed re-calibration of the AXAF Guide and Acquisition Star Catalog (AGASC). This calibration is motivated by inaccurate values of MAG_ACA, the expected magnitude in the ACA bandpass, at B(Tycho)-V(Tycho) > 0.9 in the flight AGASC (version 1.5). This error results in predicted values exceeding half a magnitude brighter than the actual ACA observed magnitude. The below plot shows this offset. See the previous report for more details. To correct this problem, we propose re-populating the MAG_ACA and MAG_ACA_ERR arrays in the AGASC for all stars originating from the Tycho-2 catalog (about 99% of the usable acq/guide stars) using a calibration based on the V-Band magnitude and the actual Tycho-2 color.

Data Products

New predictions for MAG_ACA are determined from a database of 31219 stars acquisition attempts from the past five years of operations. Only stars that were positively identified by the OBC and having valid Tycho-2 magnitude and color information were used. This criteria reduces the number of stars used for re-calibration to 30238 ( Note that this statistic indicates a dominance of Tycho-2 stars over acquisition candidates in addition to fantastic camera performance. ). Previous calibrations have used flight data in the form of guide star magnitudes from a GRETA script. While the acquisition process uses only a single data point (the OBC magnitude at the moment the star is acquired), the GRETA data was averaged over an entire observation. It is good and useful to show that these are equivalent. Below are plots of the acquisition and guide star magnitude offsets for 1587 observations. Median values are discretized to 1/16 mag. The center of each distribution is less than 5/1000 magnitudes away from zero.

The calibration discussed here is designed to only re-calculate values for Tycho-2 stars in the AGASC. These stars make up the bulk of acquisition and guide star candidates, but do not include many of the fainter stars that act as spoilers. It is impossible to re-calibrate these fainter stars as they are rarely observed intentionally. Below is the distribution of Tycho-2 and non Tycho-2 stars with magnitude. Note that the distributions are not cumulative, the red distribution is simply plotted on top of the black one.

Analysis and Results

The expected ACA magnitude is calculated as a function of the Tycho V-Band magnitude and Tycho color (VT-BT). Below is a the observed offset of Tycho magnitude from ACA magnitude against VT-BT, as well as the fit to be used for calibration.

The data were grouped into color bins with width of 1/10th mag. Color bins containing less than 100 observations were combined with stars from the adjacent bin. To eliminate outliers that do not represent the majority of observed stars, the data are then iteratively sigma-clipped about the bin mean. This further reduces the number of data points to 29066. We then apply a cubic spline fit with seven equally spaced nodes to model the clipped data. The number of nodes was chosen to provide a best fit to the apparent ridge line of points. The data was not truncated to a maximum color before fitting, the entire range was fit for completeness. Stars with colors greater than 2.5 should be naturally selected against in SAUSAGE due to their large error terms. A close up with of the most relevant region of the data is below.

Previous values of MAG_ACA_ERR were determined by simple propagation of AGASC MAG_ERR and COLOR1_ERR by the following relation:

scolerr= COLOR1_ERR/100.
var_aca_err= MAG_ERR^2 + (scolerr^2)*[ C1 + 2*C2*COLOR1 + 3*C3*(COLOR1^2) + 4*C4*(COLOR1^3)]^2
MAG_ACA_ERR= 100.*sqrt(var_aca_err)
Where the above polynomial is the derivative of the AGASC 1.5 MAG_ACA magnitude correction.

For AGASC 1.6, the catalog error is calculated as follows:

scolerr= COLOR2_ERR/100.
var_aca_err= MAG_ERR^2 + (scolerr^2)*[fit_deriv]^2
MAG_ACA_ERR= 100.*sqrt(var_aca_err)
where fit_deriv is the numerically calculated derivative of the spline fit correction.

In addition to simply recalculating MAG_ACA_ERR, we now have the means to see how the catalog empirically matches the ACA response by looking at the spread of observed stars.

For each color bin defined above we characterize the aca response error as the mean deviation from the expected error.

obs_err = MAG_ACA - obs_mag
response_err^2 = | sum(obs_err)^2 - sum(MAG_ACA_ERR/100.)^2 | / N_bin

So for each star in the catalog the AGASC 1.6 MAG_ACA_ERR is calculated by adding the appropriate response error (depending on color only) to the propagated catalog error. Below are three scatter plots that show the effect of the above calculations. The first plot is the AGASC 1.5 MAG_ACA_ERR vs. COLOR2. Note that stars with COLOR2 > 1.76 (1.5/0.85) have errors set to 0.87 mag. The second plot shows the catalog MAG_ACA_ERROR calculated for AGASC 1.6. The third plot shows the combination of the catalog error and the response error.

Above are shown MAG_ACA_ERR values for the current and proposed AGASC. The spike in the current catalog is the result of the COLOR1 (B-V) array being truncated to B-V = 1.500 and having a null color error. The new calibration uses the COLOR2 array (BT-VT) which is not truncated. The below plot is a zoom in of the main region of interest plotted on log scale. On the whole, errors are slightly larger but should have only a minor impact on star selection. SAUSAGE testing will need to be performed to verify this before official release.

The improvement in ACA_MAG prediction is presented below. The new calibration performs accurately well past the point were AGASC 1.5 fails. For BT-VT < 1.0 the calibrations are essentially equivalent: the median residual is 0.006 mags for the new calibration and 0.010 for AGASC 1.5. For BT-VT > 1.0, however, there is no contest: the median residual is 0.005 mags for the new catalog and 0.220 for the old one.





Aspect Information main page



Comments or questions: Aspect Help

Last modified:12/27/13