Align and merge event files from different OBSIDs

Any two Chandra observations will generally use different X/Y sky coordinate systems, even if the nominal pointing of the observatory was the same. Before merging two observations one must convert the event positions of one observation to the coordinate system used by the other.
IF you trust the offset and roll accuracy of the coordinate systems supplied by the pipelines, then this is a simple offset calculation which is independent of the data (i.e. it can be done even if you have no point sources). The CXC plans to release a tool for this, dmreproject, eventually. In the mean time you might do something in IDL like is done in reproject_event_files.pro, which unfortunately doesn't preserve the FITS headers perfectly.

However, in general one will want to use a set of source positions from the two observations to derive a least-squares coordinate transformation. The IDL program register_frames.pro will do this.

• Match up the sources from the two observations. Two source catalogs from wavdetect (FITS format) can be displayed, edited, and matched using the program combine_catalogs.pro, called like this:
  
  combine_catalogs,['obsxx.sources','obsyy.sources'],XOFFSETS=xoffsets,YOFFSETS=yoffsets
  
  The optional vectors XOFFSETS & YOFFSETS can be used to supply rough offsets between the catalogs so that matching sources come our near each other. Buttons allow one to print information about each source and to remove sources that don't appear in both catalogs, or that are undesired for some other reason. If the edited catalogs are the same length when saved, then combine_catalogs will match up sources and save the catalogs in the same order (required for next step).



• Choose one observation as the "reference" frame and the other as the "skew" frame. The skew observation will have its event positions changed to match the reference frame.
  
  
• Load the X & Y positions for the sources into IDL and call register_frames.pro. The code to do this will depend on how your source list is stored and how you did the matching. For FITS catalogs you'd do something like this:
  
  c1 = mrdfits('reference.cat', 1)
  c2 = mrdfits('skew.cat', 1)
  register_frames, c1.x, c1.y, c2.x, c2.y, mask
  
  where "mask" is an undefined variable. A least-squares transformation (offset-plus-rotation) will be computed and the fit residuals will be plotted. The variable "mask" will be returned as a byte array containing 1's.
  Examine the residuals and decide if any of the fiducial points should be excluded from the fit. To exclude a point set its mask value to zero, e.g. mask[17]=0, and call register_frames again. Continue iterating until you're satisfied with the fit.