[Ifeffit] Re: still the spike

Bruce Ravel bravel at anl.gov
Thu Oct 19 13:42:34 CDT 2006


Silvio,

I think you should take a step back.  Instead of asking yourself what
you *want* your data to look like, you should ask yourself how the
algorithms used actually work.  The background removal is done using a
spline:
  http://cars9.uchicago.edu/~newville/autobk/ProgramDoc/

Splines, by their very nature, have limited ability to conform to an
arbitrary shape.  By restricting the spline to use a number of knots
determined by the maximal information content of the region between 0
and Rbkg, the spline can only have those Fourier components
(neglecting leakage due to the finite transform range).  One result of
the limited Fourier content of the spline is a difficulty following
the most swiftly varying parts of the data.

Typically the most swiftly varying part of the data is the bit near
the edge.  That has the additional issue of being the beginning of the
data range.  Consequently, that end of the spline is only constrained
to follow the data on the higher energy side.

The result of this is that the spline sometimes fails to follow the
data near the edge -- at least it fails to follow it in the same way
that you would draw the background with a crayon.  It does a perfectly
fine job of optimizing the low-R Fourier components.

So what is the impact of this spike in the background function?  Well,
instead of gnashing our teeth, let's check it out.

Make a copy of the HGSREDPOW_026.001 group.  For both of them use the
same background removal parameters you currently have in your project
file.  Also for both of them, set the Fourier tranform dk value to 0
-- that is the value used in the background removal.  For the origin,
set the FT k-range from 0 to 13.5 -- i.e. the entire k-range.  For the
copy, set the FT k-range to [1:13.5].  Overplot the two data as
chi(R).  They are not very different, even though the copy avoids
using the "bad" region at low k in its FT.

In any case, at the level of Athena, you don't actually have
sufficient information to be critical of the fine details of the
background removal.  You can, I suppose, jigger the background removal
parameters until you have "despikified" your data to your
satisfaction.  However, the real question is whether you can analyze
and interpret your data and whether your fitting results are highly
correlated to the background removal.  Any thing else is merely
aesthetics.

B

-- 
 Bruce Ravel  ---------------------------------------------- bravel at anl.gov

 Molecular Environmental Science Group, Building 203, Room E-165
 MRCAT, Sector 10, Advanced Photon Source, Building 433, Room B007

 Argonne National Laboratory         phone and voice mail: (1) 630 252 5033
 Argonne IL 60439, USA                                fax: (1) 630 252 9793

 My homepage:    http://cars9.uchicago.edu/~ravel 
 EXAFS software: http://cars9.uchicago.edu/~ravel/software/exafs/




More information about the Ifeffit mailing list