[Ifeffit] choosing dk values

[Scott Calvin]

Hi Mike,

I would add to Matt's reply that it is sometimes quite important to realize
that for a Hanning window and similar windows where there is a "sill" over
which the window function goes from 0 to 1, the sill extends from kmin-dk/2
to kmin+dk/2 and from kmax-dk/2 to kmax+dk/2. This means that as you
increase dk you actually sample more k-space, albeit in a suppressed way.
Why is this important? Suppose you have something really screwy in your
data at k=11.9 inverse angstroms...a second edge comes to mind as a
possibility. You might then reasonably set kmax to 11 and think you're
staying away from the problem. (Well, it might be messing up your
background, but that's an issue for another day...) But suppose you make dk
= 2. Now you are actually sampling data up to k = 12, which includes the
really screwy bit. In some cases that can be a source of problems...it took
me a long time to catch on to that particular wrinkle when I was starting out!

Of course, with Artemis and Athena this is easy to catch, because you can
plot the window function by checking the box marked "Window" (that Bruce is
a clever dude).

--Scott Calvin
Sarah Lawrence College

At 01:13 PM 11/1/2004 -0600, you wrote:
>Hi Mike,
>
>On Mon, 1 Nov 2004, Michael A Groves wrote:
>
>> I'm doing some fitting with Artemis, and I noticed that, for
>> different values of dk, I get better or worse fits.  The only
>> information I've gotten so far about dk was in the IFEFFIT
>> documentation, and it basically said, "Don't mess with it."  
>> Can anyone give me more information on choosing a good dk value?  
>> Or, should I just stick with the default?
>>