In a number of cases studied, the proposal coordinates were from the printed version of the Yale Bright Star Catalog. One problem is that the equinox 2000 positions in the BSC are given in the FK4 (Besselian) reference system. The proposal system assumes that equinox 2000 and later coordinates are in the FK5 (Julian) reference frame, and that earlier ones are in the FK4 frame. This can be overridden by specifying B2000 instead of J2000 for the equinox in the proposal. The latest digital version of the BSC (BSC5) is in J2000. The 1950 edition of the SAO catalog is in B1950 (FK4), and a digital version is available for J2000 (FK5). An error of up to 1.5" can result from assuming BSC positions are J2000 instead of B2000 in the proposal.
Another common problem with target coordinates is that they lack precision. For example, in the BSC, RA is given to the nearest 0.1s and DEC to the nearest arcsecond. This can cause an error of up to 0.75" in RA and 0.5" in DEC. The SAO coordinates have higher precision, 0.001s in RA and 0.01" in DEC, and should be used when possible.
A common error source is not specifying proper motion or specifying it in the wrong units. It is critical to follow the latest version of the proposal instructions on this because the units have changed. Even small proper motions are significant at the resolution of HST images.
Residual pointing errors (after coordinate errors and aperture location changes) range from 0.26" to 1.84" . The average is 0.93" and the median is 0.86" . There are no obvious trends in any coordinate system. These are errors which cannot be accounted for by a proposer, being due to guide star position errors, FGS alignment uncertainties, and residual aperture location errors. Using Guide Star Catalog positions may help reduce the error between target and guide stars. Most of the targets used in this study were too bright to have true Guide Star Catalog positions.
In summary, a target with good coordinates (and proper motion) referenced to the SAO catalog can typically be placed within 0.9" of a specified aperture. However, errors of around 1.5" occasionally happen.
This update also affects the position information placed in the image headers, which maps sky coordinates onto each individual CCD. Observations taken before April 11, 1994 have preliminary plate scales, rotations, and reference pixel locations in their image headers. Thus, the sky coordinates associated with a given pixel will be different for otherwise identical images taken before and after April 11, 1994, due to improvements in the aperture locations. The change is primarily an approximate 3" shift, as well as a small rotation. There is a 0.8 degrees rotation for WF2, and smaller rotations for the other chips (0.28 degrees in PC1, 0.46 degrees in WF3, and 0.06 degrees in WF4).
The STSDAS task METRIC, which is used to assemble the 4 CCDs into a single image, uses this header information. Hence images taken before 11 April 1994 will not properly assemble into a single image. A new STSDAS task UCHCOORD will be available in the next STSDAS release (early 1996) to update the header group parameters to reflect the improved plate scale, shift, and rotations. This task will supersede the task UCHSCALE in the current STSDAS release, which can update only the scale, not the rotation or shift.
We also note, that in April and May 1996 two updates were made to the (V2,V3) coordinate system. This update should be transparent to most observers. The purpose was to remove a slow drift in the position of WFPC2 in the HST focal plane; the largest change was 0.6" . (See Table 3.15 on page 61
Other programs report similar 3 mas pointing accuracy if simple re-acquisitions are done between orbits. Approximately once per day a "full-up" acquisition is usually required (for engineering reasons) where the dominant FGS is fixed in position, but the sub-dominant FGS performs a spiral search for the guidestar and tracks wherever the star is found. On rare occasion these full-up acquisitions produce position errors of several hundred mas, and field rotations of up to ~0.1 degrees , relative to previous images of the same field. This may impact long sequences of exposures requiring half a day or more to execute.
7.7.1 Absolute Pointing Accuracy
We have looked carefully at a sequence of images to assess the absolute pointing performance that HST delivers to WFPC2. The apertures used in the observations studied were either PC1, PC1-FIX, or WF2. The observed positions of stars on WFPC2 images were measured and compared with the proposed coordinates and apertures. Where necessary, coordinate and proper motion errors were accounted for (with the assumption that SAO catalog coordinates are exact - they form the astrometric basis for the guide star coordinate system). The typical residual pointing error is 0.86" , with 1.84" being the largest error seen. This study did bring out several easy-to-make target coordinate errors (which we corrected in the analysis, but which frequently dominated the pointing error), so we discuss these first.7.7.2 Updates to Aperture / Coordinate Systems
We note that on 11 April, 1994 an update was made to the spacecraft database which tells HST where to place targets relative to the FGSs. This impacted both the location of targets in the WFPC2 field-of-view, and the position reference frame in the image headers. The nominal (or intended) pixel locations of the apertures in the WFPC2 focal plane did not change. Only the (V2,V3) coordinates of the apertures changed, as their locations relative to the FGSs became better known. For example, PC1 and PC1-FIX are designated to be at pixel (420,424.5). Before April 1994, this aperture was thought to be at (V2,V3) =(4.95" ,-30.77" ), which using the most current information was actually located at pixel (X,Y) = (459.8,377.3). Since April 1994, the aperture in the spacecraft's database has been at (V2,V3) = (1.87" ,-30.96" ) or, assuming the current best estimate is exactly correct, at (X,Y) = (414.9,428.1). Thus, for the same coordinates and aperture, the pixel position of a target in an image taken before April 1994 could be nearly 3" different from its position in later images, due to aperture updates. Similar corrections apply to all WFPC2 data taken before this date.7.7.3 Pointing Repeatability
The Hubble Deep Field (HDF) afforded an opportunity to study the repeatability of pointing over many images and acquisitions of the same field. The pointing appears to have been stable to better than 5 mas accuracy while taking many images of the same field without interruption over several orbits. The accuracy for full-up acquisition of the same field after slewing to other targets appeared to be ~10 mas typically, with occasional 20 mas errors seen. However, a few large errors were seen; in about 1 in 100 acquisitions the FGSs locked-up incorrectly resulting in a ~1" error. 7.7.4 Tracking Modes
Three guiding modes are available: (1) Gyro Hold, (2) Coarse Track, and (3) Fine Lock. Fine Lock (PCS MODE FINE) is used by default, since use of Coarse Track may be harmful to the Fine Guidance Sensors. Use of Gyro Hold (PCS MODE GYRO) is not generally recommended, even for snapshot (SNAP) observations, since the pointing accuracy is only 14" . Also the drift rate is 0.0014" s-1 so exposures >100s can result in smeared images. However, if the reduced pointing accuracy can be tolerated, and the exposures are only a few seconds or less, Gyro Hold can give a significant savings in the target acquisition overhead time.