skip to content
 
Suzaku Guest Observer Facility

XISSIMARFGEN: Tips for Reducing Run Time


1. Reducing the number of energy bins in ARFs

Computation time of an arf table is proportional to the number of calculating energy steps, which are set up with the "estepfile" parameter:

estepfile [filename]
   Energy step file or built-in steps. The built-in energy steps are:
   "full" : calculate effective area for each RMF energy bin. Very slow.
   "dense" or "default" : dense sampling (2303 steps). Slow.
   "medium" : medium sampling (157 steps). Moderate.
   "sparse" : sparse sampling (55 steps). Fast.

Fits of a decent spectrum do not improve significantly with an arf generated with estepfile=default compared with an arf with estepfile=medium, but arfs with the default energy step need 14.67 (= 2303/157) times longer computation time to be made than those with the medium energy step.

2. Optimizing the "num_photon" and "accuracy" parameters

When limit_mode=NUM_PHOTON, computation time is also proportional to the number of faked photons set up at the "num_photon" parameter.

For a point source ARF, limit_mode=NUM_PHOTON num_photon=400000 is recommended, but limit_mode=MIXED num_photon=200000 accuracy=0.005 is acceptable for faint sources.

For a uniform sky ARF, limit_mode=MIXED num_photon=2000000 accuracy=0.005 is recommended.

For ARFs of extended sources (including uniform sky ARFs), you should also check accuracy of ARFs with your own eyes, by plotting the effective area in xspec.

3. Running xissimarfgen with a fast CPU using a fast code

The Monte-Carlo ray-tracing simulation runs numerous floating point calculations, and so Athlon64 usually runs xissimarfgen faster than Pentium4. If available, 64-bit codes compiled on the 64-bit Linux runs about 1.5 times faster than 32-bit codes on the same PC.

4. Running arf caluculations for different sensors (XIS0/1/2/3) on different PCs

5. Generating ARFs of multiple accumulation regions in the same observation at once

6. Defining the smallest emission region possible when generating diffuse source arfs

If an emission region defined at the "source_image" parameter is too large compared with the event extraction region, computation time gets slower without improving quality of the simulation. For generating a uniform sky ARF, we recommend you to use source_mode=UNIFORM source_rmin=0 source_rmax=20.

Similarly, if the spacecraft attitude is not stable after a maneuver and the emission region goes out from the event accumulation region, the ARF calculation gets slow, too.


NEXT: Complete List of Parameters
BACK: Recommended Usage and Caveats I
UP: Xissimarfgen top page


If you have any questions concerning Suzaku, visit the Feedback form.

This file was last modified on Wednesday, 19-Sep-2007 21:18:19 EDT

NASA Astrophysics

  • FAQ/Comments/Feedback
  • Education Resources
  • Download Adobe Acrobat
  • A service of the Astrophysics Science Division (ASD) at NASA/ GSFC

    Suzaku Project Scientist: Dr. Robert Petre
    Responsible NASA Official: Dr. Andy Ptak

    Privacy Policy and Important Notices.