Re: Pile-up

[Cc'd to isis-users]

On Mon, 11 Dec 2000 18:11:54 -0500, you said:
>Hi John - I'm wondering how I can use ISIS to deal with a piled-up point 
>source spectrum since I cannot find anything about pile-up in the user manual. 
>My source was taken without grating with ACIS-S and the pile-up fraction is 
>about 10-20%. I tentatively extract the spectrum with an annulus region 
>centered on the source and it's softer a little bit by comparing to the one 
>with circular region.  Any suggestion?

Is your point source on-axis?  If so, I suggest that you extract
counts in a circular region with a 4-pixel radius.  This will get you
about 95% of the encircled energy.  You have to go out much farther to
pick up the other 5 percent,  which in my view is not worth it since
you will also pick up background.

Then in isis issue the appropriate commands to load the data, e.g.,

   () = load_data (data);
   () = load_arf (arf);
   assign_arf (1,1);
   assign_rmf (rmf, 1);

Now you will have to binup the data.  I use the command:

   group_data (1, 12);

to group every 12 pha channels together.  Here, group_data is a
function that I defined--- see below for its definition.

Then specify a model, e.g.,
   
   import ("xspec");
   fit_fun ("wabs(1)*powerlaw(1)");

Now you have to indicate that you want to fit using pileup:

   set_kernel (1, "pileup");

Finally, edit the parameters:

   edit_par;
   list_par;

This will produce something like:

  1  wabs[1].NH_22       0     0              0.01         1e-5  0.1
  2  powerlaw[1].norm    0     0             0.003         1e-5  0.01
  3  powerlaw[1].alpha   0     0                 1            0  0
  4  pileup.<1>nregions  0     1                 1            1  1.5
  5  pileup.<1>g0        0     1                 1            0  1
  6  pileup.<1>alpha     0     0               0.5            0  0.75
  7  pileup.<1>psffrac   0     1              0.95          0.9  1

The pileup<1> parameters are specific to the pileup model.  For an
on-axis point source, allow only alpha to vary.  The leave the psffrac
value fixed at 0.95.  This indicates that 95 percent of the
encircled energy in the extraction region fell within the central 3x3
pixel.

Finally, run the fit:

   fit;


I would do it several times, starting from random parameter values.
The chi-square space is quite complicated.

Here is my group_data function.  I have it defined in my .isisrc file:

public define group_data (ds, num)
{
   variable idx, len;

   rebin_data (ds, 0);		       %  unbin data

   (idx,,,) = get_data (ds);
   len = length (idx);
   
   idx = Int_Type [len];
   
   variable i, s;

   s = 1;
   _for (0, len-1, 1)
     {
	i = ();
	!if (i mod num)
	  s = -s;
	
	idx[i] = s;
     }

   rebin_data (ds, idx);
}


----
You received this message because you are
subscribed to the isis-users list.
To unsubscribe, send a message to
isis-users-request_at_email.domain.hiddenwith the first line of the message as:
unsubscribe