Hi Johannes,

I guess I hadn't tested the Chantler results below about 1 keV, and certainly not at 100 eV. To be clear, Chantler claims the error bars on these value of "50 to 200%" below 200 eV.   But the spline interpolation is much, much worse than that!

Spline interpolation is definitely preferred at higher energies, and where the grid of tabulated data is fine enough.   But we should better explore when that is valid, and/or automatically use linear interpolation below some energy, and/or allow the user to select the interpolation order.

Anyway, thanks very much.   We should make this better at all energy ranges.

--Matt


On Wed, Oct 14, 2015 at 1:11 AM, Dr. Johannes Zellner <johannes@zellner.org> wrote:
 
Hi,
 
the larch xraydb plugin doesn't provide correct numbers for chantler f1 data around edges.
Apparently this is due to incorrect interpolation in xray/xraydb.py in line 274.
Below is a plot of "delta" from xraydb_plugin.xray_delta_beta() for Silicon which is terrribly wrong.
 
 
The more correct "delta" as from chantler is shown in the following plot:
 
 
--> I'd recommend to replace (line 274 in plugin xray/xraydb.py):
 
out = UnivariateSpline(te, ty, s=smoothing)(energy)
 
by a simple linear interpolation
 
out = np.interp(energy, te, ty)
 
The latter yields linear segments between chantler's data, but this is still better than the totally unphysical spline interpolation.
 
Best regards,

-- 
Johannes

_______________________________________________
Ifeffit mailing list
Ifeffit@millenia.cars.aps.anl.gov
http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit




--
--Matt Newville <newville at cars.uchicago.edu> 630-252-0431