Bright Star Astrometry Issue

Microtest results

STS astrometry

Reprocessing STS.2010

The STS astrometry showed offsets in the positions of stars in about half of the chips, with mean offsets of about 0.8 pixels. Gene first characterized this, and showed that changing the stars used for the astrometry from the brightest non-saturated stars to fainter stars improved the agreement for the majority of stars, although it created a trend of bright stars having larger position offsets, as shown in the following figures (x is the reference catalog magnitude, y is the shift in RA in arcseconds).

The original thought was that these offsets were due to burntool oversubtraction damaging the star and skewing the centers, or that the STS observing method was promoting persistence effects. However, on closer examination, no obvious persistence effects were seen, and the small offsets in the pointing prevented stars from directly falling on the same pixels on subsequent visits to that pointing. In addition, burntool only modifies the very brightest stars, with no stars fainter than m_instrumental = -14 having any correction applied. The trend continues fainter than this. Running the same exposures through processing with burntool disabled shows the same effect, so the shifts cannot be caused by anything that burntool is doing to the image. A final issue pointing to a different source is the fact that RA aligns with the x-direction of the detector in these exposures. Burntool operates along columns, making it unlikely to create a horizontal shift in a star position.

2mass comparison

The SMF files for two exposures from the galactic plane (o5377g0572o and o5407g0385o) and one from another dense star field taken more recently (o5636g0041o) were matched against the 2mass catalog, and the shifts between the GPC1 and 2mass positions analysed as a function of magnitude. The following figures show dRA, dDEC, from the catalogs, along with these offsets transformed using the platescale and position angle into dx and dy coordinates. The flip in direction between OTA34 and OTA42 is due to the different orientations of the detectors. The x-axis in all figures shows the instrumental magnitude of the star.

o5407g0385o/OTA34 o5407g0385o/OTA42 o5636g0041o/OTA34

The lack of any trend in dy illustrates that this shift occurs along the rows of the detector. This is further confirmed by the large rotation (pos_angle = 43.745338) for o5636g0041o which shows the trend in both RA and DEC, but after transformation, only in dx.

A scan of all OTAs shows that this shift occurs only among 2-phase devices:


Not all 2-phase devices show the effect:

OTA24/dx OTA36/dx

Checking the differences in positions shows that in the raw cell coordinates, a bright source in GPC1 is at a smaller x position than the 2mass position. This means that an example object in cell00 has been shifted in the direction of cell10. The direction of the shift in cell coordinates is the same on all OTAs. There is some evidence of asymmetry in the raw images as well that supports this shift:

This effect starts around instrumental magnitude of -13. Using the SMF file to convert puts the peak flux for these stars around 5000 counts:

Direction of shift

The above figure shows cell00 and the full chip mosaic for OTA34 and OTA42 of o5407g0385o. The full chip mosaic has the WCS overlaid. I am ignoring DEC, and pretending that RA and x are parallel for simplicity.

  • As shown above, OTA34 has RA_IPP - RA_2MASS < 0. I've scribbled a red arrow on the chip mosaic frame illustrating the direction of this shift (from 2MASS to the GPC1 position).
  • Due to the orientation of the OTAs, OTA42 has RA_IPP - RA_2MASS > 0. I've scribbled a similar red arrow on that frame.
  • Comparing to the raw cell00 frames, the shift arrows show that for both OTAs, the IPP measurement is closer to x = 0.
  • Returning to the chip mosiac, this direction is towards the adjacent neighbor cell, cell10.