Anyone who has used the pileup model in XSPEC should be aware that a bug
has recently been fixed. Working with John Davis, who developed the
algorithm we used, we have established that the XSPEC and ISIS pileup
implementations now give identical results. There was an error in the
XSPEC implementation when three or more events were piled. The fix can
be downloaded from the XSPEC bugs page
http://xspec.gsfc.nasa.gov/docs/xanadu/xspec/bugs.html
Note that the pileup model cannot be used blindly. The fit statistic space
is complicated and it is easy to get stuck in local minima and not find
the true best fit. In particular, it is important to get the model
normalization about right before starting the fit. The normalization delta
value (the second number that can be given when defining the model or
resetting the parameter value with newpar) should then be set to a few
percent of the normalization.
Keith Arnaud
This archive was generated by hypermail 2b29 : Tue Feb 12 2013 - 01:00:10 EST