On Fri, 12 Feb 1999, Nita DRAGOE wrote:
> I wrote a graphical part for my old program Powder and peak hunting is
> indeed the reason for this Ka2 stripping. The problem I have is that I
> have some oscillations at the end of every peak after applying the
> Rachinger algorithm. I know that this is "normal" it is explained in
> the book of Klug and Alexander.
> So, after Ka2 stripping doing a peak hunt gives these additional
> oscillations as well.
> I was wondering if there are any better algorithms...
At my opinion, the solution is
-no Ka2 separation
-Instead of, use sophisticated peak hunting algorithms.
Since 1982, I have developed EFLECH, a peak hunting anf fitting program.
See
http://www.bgmn.de/related.html
Peak hunting is not done by simple maximum search in the difference
curve. Instead of, hundreds/thousands of simulated peaks (learned or
fundamental parameters) are correlated with the difference curve. The
best peak is added to the peak list, and a least square run follows.
Again, a correlation search for a single new peak follows etc.
Of course, doing so is time-expensive. Hundreds of least square runs, each
with some dozens steps, must be performed. Total time is doubled by
the correlation search. But, time consumption is reduced by two facts:
-Powerfull modern PCs
-The whole diagram is automatically divided into parts. The
search/fit algorithm is performed on every part.
Doing so, a whole diagram of much more than hundred peaks is
solved in about some hours.
J. Bergmann
[EMAIL PROTECTED]
