HEAD 2002 Pileup Talk  

Data Preparation and Caveats


Before one can actually use the pileup model, one must first filter the data appropriately and then construct PHA files, RMFs, and ARFs.

Do not use level 2 files!

The most important thing to note is that you must start with level 1 files and not with level 2 files. The main reason for this is that level 2 files have had so-called afterglow events removed. For bright sources this means that 10 to 20 percent of the events from the source have been misidentified as cosmic ray events and removed from the level 2 files. Hence, you have to go back to the level 1 files. See the CIAO acis_detect_afterglow thread for information about how to keep afterglow events.

Exercise care in creating light curves

The pileup model assumes that photon arrival times are Poisson distributed. This means that lightcurves must be constructed to determine regions where the incident source flux is fairly constant. In doing so, one might be tempted to use events centered upon the source itself for the construction of the lightcurves. However, this is a bad idea since for a bright source, the relationship between the source flux and the count-rate is non-linear. Instead, use either the events in the wings of the PSF, or the events in the ACIS readout streak. To illustate this point, here is a lightcurve of II Peg (obsid_1451) constructed from the source events:

II Peg light curve using source events
From this one would conclude that about 30000 seconds into the observation, the source flux increased by about 10 percent. However, look at the lightcurve constructed from the events in the ACIS readout streak:
II Peg light curve using streak
(The points in red represent the background lightcurve). The flux actually increased by about a factor of 3!

Use the right extraction region

For an on-axis point source, it is recommended that a circular region with a 2 arc-second radius be used (4 ACIS pixels). Such a region encompasses 95 percent of the incident flux and one would have to go out to a much larger radius to get the remaining 5 percent at the risk of decreasing the signal to noise ratio. Schematically, the recommended extraction looks like
Figure showing the extraction region
Also shown at the center of the figure is the 3x3 event detection cell where pileup is assumed to occur. The event detection cell is expected to enclose approximately 90 to 95 percent of the flux falling into the full extraction region. This sets a constraint on the pileup model's psffrac parameter to be between 0.9 and 0.95.

Do not use a canned ARF

Use mkarf to construct the proper ARF that is appropriate for your observation. In addition to being easy to use, mkarf takes observation-specific dither and bad pixels into account, and writes a value representing the fractional exposure time lost to dithering over bad pixels to a keyword called FRACPROB. The value of the keyword is picked up transparently by isis, but sherpa requires the user to set it explicitly.

Get the EXPOSURE right!

The pileup model uses the frame-time and exposure time to compute the number of CCD frames. If the value of the EXPOSURE keyword in the ARF or in the PHA file is wrong, then the number of frames will be wrong, causing the pileup model to fail. It has been our experience that a mkarf generated ARF has the correct value for the exposure keyword, whereas the value in the PHA file produced by earlier versions of CIAO was often incorrect. For this reason isis was coded to use the value of the EXPOSURE in the ARF and ignore the value in the PHA file. However, both sherpa and xspec use the value in the PHA file. In any case, check the value of the EXPOSURE keyword in both the ARF and the PHA file and correct them if necessary. (HINT: If you are using isis, you can also use the set_arf_exposure function to correct the exposure after the ARF has been read in. I am not aware of any such feature in the other two programs.)

Previous: Pileup Model Next: Spectral Fitting Software

This page was last updated May 31, 2002 by John E. Davis. To comment on it or the material presented here, send email to davis@space.mit.edu.
Valid HTML 4.01! Made with JED Viewable With Any Browser