In January 2012 it was discovered that because of
aim point drift
over time, recent observations using the HRC-S (whether using a grating or not)
may have a small fraction of source (or 0th order) events that fall
on the CRSV=100 tap. (Read
here for a short explanation
about taps and the degapping process.)
The current degap map
(hrcsD1999-07-22gaplookupN0002.fits, from 2005/11/16), which converts
event positions from RAW to CHIP (which are then converted
to DET and TDET, and ultimately to SKY) has residual errors
for that tap, leading to asymmetry and broadening of the PSF.
It does not affect the positions of events dispersed
by the LETG, although the wavelength calibration may be slightly affected
(see below). Figures 1-3 show examples from ObsID 13025, a
Calibration observation of HZ43 made in March 2011, and one of the
|Figure 1. 0th order in SKY coordinates. Note that the PSF is extended toward the upper right.||Figure 2. 0th order (extracted with 30-pixel-radius circle in SKY coordinates) displayed in CHIP coordinates, showing the dither pattern. The dither pattern on tap CRSV=100 appears fuzzier than elsewhere and there is a significant gap between taps 99 and 100.|
|Figure 3. 0th order in SKY coordinates using only events with CRSV=98:99 (left) or CRSV=100 (right). The green circle (5-pixel radius) is centered on the same pixel in each.|
This problem started to show up in late 2010 as the dither pattern of sources at the aim point began to drift onto CRSV=100. We are aware of 10 GO/GTO observations of point sources that have been affected: ObsID 12154, 12203 (just barely affected), 12332, 12324, 12325, 12401, 12916 (probably negligible), 13184, 13240, and 13651. There is no guarantee, however, that this list is complete. Observers concerned about the source/0th order PSF should check any observation made since the beginning of 2010 following the procedures outlined below. In addition, any observation made since the beginning of the mission that observes a field with multiple or extended sources may be affected.
One way to tell if your observation is affected is to extract an image of the
target of interest in SKY coordinates from the Level 2 event file (*evt2.fits)
using dmcopy and then display
those extracted events in CHIP coordinates using
ds9 "extractedsource.fits[binkey=(chipy,chipx)]" &
This produces an image like that in Figure 2. Any events with CHIPY between 9140 and 9395 fall on the tap with CRSV=100. (Within ds9, with Bin/Binning Parameters/Bin Filter, one can reduce the background and improve image contrast using pi=50:150,pha=0:254. This filtering can, of course, also be applied to the extracted file using dmcopy.)
Another method is to quantify the fraction of events falling on the
problematic tap using
dmstat "extractedsource.fits[chipy=9140:9395][cols sky]"
dmstat "extractedsource.fits[cols sky]"
Position errors only occur along the long axis of the HRC-S (in detector coordinates RAWY and CHIPY, and grating coordinate TG_R). Because of the roll angle, however, position errors will generally occur in both X and Y SKY coordinates.
An improved degap map for regions near the aim point is currently under development and a calibration observation to map out all of the CRSV=100 tap has been scheduled for April 2012. In the meantime we provide the following recommendations for observers and users.
We anticipate that the revised degap map will be ready in time for all future HRC-S observations. In the unlikely event that a new degap map can not fix the problem, observers who need optimal resolution around the aim point can use a small pointing offset, e.g., Yoffset=-10", to avoid tap 100.
Timing and event rate studies will be unaffected provided that
the source extraction region is enlarged to include misplaced source events.
If analysis focuses
on source morphology, however, users may want to exclude data from CRSV=100.
Level 1 event files can be filtered using
dmcopy "evt1.fits[crsv!=100]" new_evt1.fits
Both Level 1 and Level 2 event files--the latter do not carry the CRSV
column--can be filtered using
dmcopy "evt2.fits[chipy=:9139,9396:]" new_evt2.fits
The above filtering yields an event file with good event positions, but does not adjust GTI block data to account for the missing events. This means that light curves, exposure maps, effective areas, etc. will behave strangely and/or no longer be appropriate for the filtered file. Although it is possible to analyze the filtered data and correct for the time-dependent effects of missing events (contact the Helpdesk for assistance), it is strongly recommended that users wait for the new degap map and associated reprocessing instructions before attempting to obtain reliable analysis results.
For most LETG/HRC-S observations, 0th order is of no importance other than centering the spectral extraction regions and defining the origin of the dispersion coordinates. Event position errors for CRSV=100 can lead to a distorted 0th order PSF, which may lead to an error in the centroid used to define the 0th order position. This error could be significant if the user is interested in using the spectrum for accurate line profiles and absolute wavelengths.
In Figures 1-3, the net centroid error is roughly one half pixel. Most affected observations have a smaller fraction of source events with CRSV=100 so centroid errors will be smaller, but a few observations will have larger errors. An error of 1 pixel in centroiding 0th order along the dispersion axis (tg_r) will produce an error of 0.0074 Å in wavelength. The discrepancy between + and - orders will be double this, or 0.0148 Å. For comparison, the FWHM resolution of the LETG is about 0.05 Å. If + and - orders are combined, the resulting summed spectrum will have lines broadened by approximately this amount but their absolute wavelengths will be unaffected. Centroiding errors in the cross-dispersion direction (tg_d) are believed to be negligible.
If an observer wishes to obtain spectra with the the best possible
wavelength calibration he must filter out the
CRSV=100 events from the Level 1 event file by following the
procedure described above
(dmcopy "evt1.fits[crsv!=100]" new_evt1.fits);
the remaining data will yield
an accurate determination of the 0th order location.
The new Level 1 event file can then be reprocessed following the
steps in the
Grating Spectra thread. Note that using the filtered 0th order
data for anything other than anchoring the wavelength scale is not
advised; see the caveat in the previous section,
"Past Observations without a Grating."
Last modified: 02/17/12[an error occurred while processing this directive]