[Xrays@aps.anl.gov] (no subject)

[Jon Tischler]

The point that could have the greatest impact would be to get the  
written is a manner that is uniform across the facility.  This way,  
analyzing data, particularly the immediate analysis during an  
experiment could happen more efficiently.  Right now there are  
multiple programs that take data, all in their own fashion, and even  
for one program, there is no consensus on how to use it to save data  
(e.g. spec is used multiple ways here).

Simple things, such as naming the channels in spec (is incident beam  
intensity called, Io, I0, iO, io, ion, ionc, ...?).  Is the scattered  
beam detector (scint, det, bicron, Bicron, detector, oxford, signal,  
NaI, ...?)  Is the detector a photon counter (which needs a dead time  
correction) or is it an integrating detector?  Even little things  
like this can bring some sanity to the analysis.  It is not enough to  
have a description of the detectors, but a dependable naming (or some  
means of identification) is needed so that at analysis time the  
computer will know what it is.  People will spend weeks writing  
routines to operate an experiment, and only minutes on a way to write  
the data  This lack becomes evident when it is time to analyze the  
results; especially when something is added or changed in the data  
file.  I've even heard that some people are still using two column  
files without any headers, because it was once easy to write.  I  
guess they deserve what they get.


Continued support for EPICS is essential, it is one of the few things  
that can be depended upon to work in more than one experiment.


EPICS is good for controlling things, but is no good at taking data.   
Should the issue of data collection be revisited?


Jon