Subsections

• 7.17.1 Actions
• 7.17.2 Usage and Inputs
• 7.17.3 Data Products
• 7.17.4 Choosing Among Position Estimates
• 7.17.5 PSF and Neighborhood Image Binning

7.17 CHECK_POSITIONS Stage

This stage is run once, after trial extractions have been performed for each observation and the MERGE_OBSERVATIONS stage has been run, as shown in Figure 8. The source position estimates produced by this stage may be used to re-position the sources using the tool ae_source_manager (§7.2.2).

7.17.1 Actions

• A multi-ObsId data image of the source's neighborhood is constructed by reprojecting the neighborhood events from each observation, filtering the events in the energy range specified, then binning them into an image using the same bin size as the multi-ObsId PSF.

• An estimate of the source position (§5.3) is obtained by correlating the multi-ObsId data image and the multi-ObsId PSF (§7.16), and is saved as keywords RA_CORR and DEC_CORR in source.stats.

• The multi-ObsId data image is reconstructed using a simple Maximum Likelihood algorithm (max_likelihood.pro in the IDL Astro Library).

• An estimate of the source position (§5.3) is obtained by centroiding a 3x3 pixel neighborhood around a nearby peak in the reconstructed image, and is saved as keywords RA_ML and DEC_ML in source.stats. Currently no uncertainty estimates are available for the correlation (RA_CORR, DEC_CORR) or Maximum Likelihood (RA_ML, DEC_ML) position estimates.

• An atlas showing the raw and reconstructed neighborhood images is produced.

• Plots summarizing the results of CHECK_POSITIONS are presented. These plots are diagnostic tools that help you spot problems in your data, in your catalog, in your CIAO configuration, in the CIAO software, or in the AE software. Some plots may also convey scientifically interesting information about your sources.

7.17.2 Usage and Inputs

acis_extract, srclist_filename, /CHECK_POSITIONS, ENERGY_RANGE=[energy,energy], MAXLIKELIHOOD_ITERATIONS=value, /SKIP_RECONSTRUCTION, /SKIP_CORRELATION, THETA_RANGE=[value,value],

acis_extract, srclist_filename, /CHECK_POSITIONS, /PLOT

EXAMPLE:
idl $\vert\&$ tee check_positions_1875.log

acis_extract, 'all.srclist', /CHECK_POSITIONS

acis_extract, 'all.srclist', /CHECK_POSITIONS, /PLOT

• The optional ENERGY_RANGE parameter defines the energy filter used in construction of the multi-ObsId data image; default is [0.5,8.0].

• The optional THETA_RANGE parameter can be used to select a subset of sources for which this stage will be run, based on their off-axis angles.

• The optional keywords /SKIP_RECONSTRUCTION and /SKIP_CORRELATION can be supplied to speed up this stage by skipping the data image reconstructions and/or skipping the computation of a correlation position estimate.

• The optional keyword MAXLIKELIHOOD_ITERATIONS can be supplied to specify the number of reconstruction iterations; default is 400. If a vector of integers is supplied (e.g. [10,20,100,500]) then the reconstructed image is saved after each of the specified number of iterations.

7.17.3 Data Products

For each source the following files are produced:

{sourcename}/source.stats:

a FITS file containing keyword information about the source

{sourcename}/neighborhood.img:

the multi-ObsId (all observations) data image (primary HDU), and reconstructed image (first extension)

mugshots.ps:

a PostScript atlas containing the raw and reconstructed neighborhood images printed 6 to a page. If you want to preview this atlas you may find that converting to PSF (distill) and viewing with Acroread gives a much better rendering than viewing the PostScript with gv or ghostview.

{sourcename}/evt.reg:

a ds9 region file containing a ``point'' region at the position of each event in neighborhood.img

{sourcename}/extract.reg:

a ds9 region file containing the source extraction polygons for all the observations, plus markers for the three estimates of the source position (catalog position, correlation peak, and mean of event data).

The following interactive plots are produced:

``Vector from Catalog to ML Position''

a vector-style plot showing the vector offset from each source's catalog position to its correlation position

``Catalog to ML Peak'', ... vs ``Average Off-axis Angle''

a scatter plot of the offset between three estimates of the source position (catalog position, correlation peak, and mean of event data) as a function of off-axis angle (averaged over all observations)

distributions of ``Catalog to ML Peak'', ...

the distributions across the catalog of the three position offsets above

``Catalog to ML Peak'', ... vs ``Source #''

a plot of the three position offsets above for each source

The data vectors plotted above are retained in an IDL save file named check_positions.sav.

7.17.4 Choosing Among Position Estimates

The task of choosing, for each source, which of the available position estimates is best has no simple or obvious solution. The mean data (RA_DATA, DEC_DATA) and correlation (RA_CORR, DEC_CORR) position estimates can clearly be corrupted by a sloping background caused by the wings of a nearby source. We strongly suggest in our personal AE recipe (§7.1) that the observer visually review the position adjustments proposed by AE before adopting them.

Franz Bauer has investigated the various techniques for calculating source positions, using multiple data sets for the Chandra Deep Field North, which has good radio and optical source positions. Based on his work, the PSU ACIS group has decided to use centroid positions for sources $< 5$' off-axis and matched-filter (correlation) positions for sources $> 5$' off-axis.

For some crowded sources the observer may feel that the original position estimate is the most reliable one, and will want to avoid re-positioning those sources. The observer is encouraged to use the PROVENANCE source property (§7.2.2) to tag such sources so they can be excluded from re-positioning operations.

7.17.5 PSF and Neighborhood Image Binning

Note that the pixel size used in the multi-ObsId PSF image file and the neighborhood image file will often differ. The multi-ObsId PSF image is saved at high resolution (by MERGE_OBSERVATIONS), but is re-binned (locally in this stage) to a more appropriate resolution before constructing a matching neighborhood image and performing image reconstruction. The IDL code below demonstrates how you would manually rebin the PSF to match a neighborhood image prior to running manually a maximum likelihood reconstruction. You'll need a TARA release dated April 2007 or later to get the proper version (1.7 or higher) of the tool maxlik.pro.

  ; To test reconstruction set data_name=psf_name
  psf_name  = 'source.psf'
  data_name = 'neighborhood.img'

  ; Determine the ratio of PSF and data bin sizes.
  psf =readfits(psf_name,  psf_hdr)
  data=readfits(data_name, data_hdr)

  rebin_ratio = sxpar(data_hdr,'CDELT1')/sxpar(psf_hdr,'CDELT1')
  
  
  ; If rebin ratio is not an integer then this approach won't work.
  help, rebin_ratio

  rebin_ratio = round(abs(rebin_ratio))

  .run
  if (rebin_ratio GT 1) then begin
    ; Rebin the PSF and save to a temporary file.
    print, 'rebinning the PSF'
    temp_name = 'temp.fits'
    cmd = string(psf_name, rebin_ratio, rebin_ratio, temp_name, F="(%'dmcopy ""%s[bin #1=::%d,#2=::%d]"" %s')")
    print, cmd
    spawn, cmd
    psf_name = temp_name
  endif
  end


  ; Now we can do reconstructions:
  Niter = 200
  maxlik, data_name, psf_name, Niter, maxlik_img, maxlik_header, /plot

  ; The final iteration is returned in the IDL array "maxlik_img".
  writefits, 'maxlik200.fits', maxlik_img, maxlik_header