From: Walter Stafford <STAFFORD@bbri.harvard.edu>
  To  : rasmb@bbri.harvard.edu
  Date: Wed, 12 Aug 1998 18:05:29 -0400 (EDT)

truncation of XL-I data and DCDT

Hi RASMBers,

	I have recently had several emails from people who are having 
trouble using the dcdt program on the Macintosh. This problem brings to light 
several considerations when using dcdt for the analysis of interfernce data in 
general.

The specific problem with the Mac version of dcdt is an "Array Out of 
Bounds" error that occurs when the operator of the XL-I selects a subset of
the full array of data. The Mac version of dcdt expects and must have all
2048 pixels to do its analysis. 

The general problem is one of correcting the fringe patterns for 2 effects. 
The folks at Beckman apparantly have been advising users to collect data 
from just inside the meniscus to the bottom of the solution column. The 
problem with this method is that it introduces serious random errors into the 
results. The newest version of Beckman's software from dcdt analysis has 
taken care of this problem (I have been told) and will be available soon.
Those of you who are using older versions cannot get correct results from
those versions. You must use either the DOS version of dcdt or the Mac
version of dcdt, or upgrade to the newest version that runs under Origin.

Details: The problem is two fold and is explained briefly as follows. The 
first problem is that there are random variations that can be quite large
in the vertical displacement of the fringe patterns from one scan to the
next. There are several reasons for this but it is inherent in the
stability of the XL-I optics. These random variations can only be removed
correctly by aligning the patterns in a region where there can be no
changes in the patterns due to sedimentation. In order to do this, data
must be taken all the way from the top of the cell so that one can use the
air-air space above the meniscus for this alignment. Both the Mac and PC
versions of DCDT can do this alignment. In the Mac version, the users are 
asked if they want to do a "Fine Adjust".

The second problem is that the algorithm that converts fractional fringes
to full fringe displacements can sometimes jump an integral nmber of
fringes if the signal is low; this occurs very often at the meniscus
position. The result is that the fringe pattern in the solution column can
differ from one frame to the next by an integral number of fringes. The
second alignment step allows the user to select a region of the solution
column at which to correct for these jumps. Just inside the mensicus is
usually good place unless a rapidly moving boundary is pulling away from
the meniscus; in that case, a spot in the plateau should be used. The
algorithm adjusts the fringe patterns by integral fringe jumps, and, so as long
as no large amounts of sedimentation occur between frames, they will be
adjusted correctly. 

Bottom line: Take all the data over the entire CCD array = all 2048 points.
Do not truncate the data!