file: /nfs/spectra/d2/GtoHAK/CasA/pileup_020104/README.txt


Purpose
-------
  Estimate the event rate per 3x3 pixel island in an ACIS image
to assess pileup level.
  (Here we do it for the zeroth-order image of the HETGS Cas A
observation, obsid 1046.)


Procedure summary
-----------------
  Make an image using all events (i.e. unfiltered
by grade, etc.  For large files the first 1 million events, say, 
can be used, "#row .le. 1000000" as long as the resulting TIME or 
EXPOSURE is correct for the shortened event list).
  Boxcar smooth the image by a 3x3 boxcar.  The values at each pixel
in this smoothed image are then equal to:

   (Number of events in local 3x3 region centered on this pixel) / 9

  Find the value above expected if there were 1/10 event per frame per
3x3 island.  Actual values well less than this value are relatively
unpiled; values approaching or exceeding this have relevant pileup.
This "10% pileup" number is given for the observation by:

   "10% pileup" number =  1/9 x 0.1 x number of frames 
                       =  1/9 x 0.1 x (max(TIME)-min(TIME))/TIMEDEL


Example on Cas A: HETGS Zeroth-order image
-------------------------------------------

- Select the first 1 million events from the evt1 or evt1a file:

[unix] fselect
Name of FITS file and [ext#][...] acisf01046_000N001_evt1a.fits
Name of output FITS file[...] acisf01046_000N001_N1M_evt1a.fits


- Use IDL to find the true time range and number of frames in this
subset of events:
IDL> evts = mrdfits('acisf01046_000N001_N1M_evt1a.fits',1)
MRDFITS: Binary table.  26 columns by  1000000 rows.
IDL> print, min(evts.time)
   1.0719868e+08
IDL> print, max(evts.time)
   1.0720491e+08
IDL> print, max(evts.time) - min(evts.time) 
       6232.4773
IDL> print, (max(evts.time) - min(evts.time)) / 3.24104
       1922.9869

There are 1923 frame times in this set of events.

The "10% pileup" number when 3x3 smoothed is thus:

             1/9 x 0.1 x 1923 = 21.3667


- Create an image of the desired region

[unix] dmcopy
Input dataset/block specification (): acisf01046_000N001_N1M_evt1a.fits[events][bin x=3700:4520,y=3770:4320]
Output dataset name (): acisf01046_000N001_N1M_img.fits


- 3x3 boxcar smooth the image

[unix] fboxcar datatype = e clobber = yes
Input FITS image to be fit[...] acisf01046_000N001_N1M_img.fits
Output FITS image[...] acisf01046_000N001_N1Mbc3x3_img.fits
X dimension of box[3] 3
Y dimension of box[3] 3


- Use ds9 to display the smoothed image and look for values that
are near or above the 21.4 value calculated above...

[unix] ds9 acisf01046_000N001_N1Mbc3x3_img.fits &

Used color settings of "invert" and "sls" and "print"ed the
image to a .ps file, e.g., CasA_zo_pileup.ps .

Using the cursor to inspect pixel values corresponding to the
colors in this image show that:
  Yellow is at about a value of 8.5, or a "4% pileup" rate.
  Orange and Red regions are less than this level and so
are at most only slightly piled up.
  Green has values around 13-14 and Blue to Black regions have
values of 17 to 23.9 and so they are roughly at the "10% pileup" level
(of 21.4.)

Note that this example is for the HETGS zeroth-order: the non-grating
Cas A observations will have pileup rates of 30% or higher in
the high-rate regions!