[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[sdpd]Re:Automatic peak position hunting



> Hi,
>
>It seems that automatic peak position hunting programs
>like PowderX, EVA 5.1, Winplotr... are unable to
>distinguish two adjacent peaks when the maximum of
>these 2 peaks are separated by only one step, even if
>the 2 peaks are clearly separated by eyes, seeing а 
>M-like shape at the peak top like this :
>
>                             |\/|
>                             |  |
>           _________|  |__________
>
>Comments ?
>
>Armel le Bail
>http://www.cristal.org/course/
>
>PS- OK we can use the cursor and do it manually,
>but the question is why automatism fails ?

Dear Armel,

Programs for automatic peak detection usually
make use of the Savitzky-Golay procedure. This
procedure is based on fitting a polynomial to the
profile and may distort the peak top seriously if
large convolution range is used. The peak search
program may automatically choose large convolution
range in order to avoid (smooth) false peaks resulting
from noise.
I tried to apply the S-G procedure to the profile from 
http://www.cristal.org/sdpd/indexrr.zip 
and found that it could resolve these two peaks if the
convolution range of 5 points was used with 2-nd degree
polynomial or 7 points with 4-th degree polynomial.
Increasing the range to 7 points with 2-nd degree or to
9 points with 4-th degree polynomial resulted in reducing
noise but made these two peaks unresolved.

An alternative procedure providing better peak
resolution with more optimal convolution range is
described in:

Solovyov L.A. Profile Smoothing and Differentiation with 
Reciprocal Polynomials. Materials Science Forum, 1996,
V.228-231, p.161-164.

Applying to the discussed doublet the Rec.Pol. procedure
allowed to resolve the peaks using 9 and even 11 points range
with 4-th degree polynomial.

Best regards,
Leonid Solovyov

-------------------------- eGroups Sponsor -------------------------~-~>
eGroups eLerts
It's Easy. It's Fun. Best of All, it's Free!
http://click.egroups.com/1/9698/1/_/40740/_/977469411/
---------------------------------------------------------------------_->