Excerpted email from Phil Massey (Sept 28, 2005)
--------------------------------------------------------------------

As I wrote a few days ago, there is a significant difference in the  
dome flats (provided as standard calibration by SMARTS) and the sky 
flats that observers occasionally succeed in  obtaining.  The difference 
amounts to 6-7%---quite significant if one intends <1% photometry!  The 
two  agree to 1% for a 2000x2000 region centered at pixel (1500,1500) 
but the two diverge quickly outside this region.

A couple of nights ago the 1-m observer, David James, kindly took the  
time to offset a star from (1500,1500) to the upper right (3500,3200), 
as a follow-up to what I had found in  comparing the dome flats and 
sky flats.  I was therefore able to answer which gave the "right" answer.

The night (050925) was photometric, and the FWHM of the images were  
a close match (3.9 and 4.1).  I used a circular aperture of radius 
5 pixels for the photometry; similar results are found for 8 and 10 pixels.

The answer, unsurprisingly, is that the sky flats  are right.  Here  
are the instrumental magntiudes:
                                                      
reduced using dome flat          reduced using sky flat
Sweet Spot (1500,1500)              
13.888                                     13.883
Upper Right (3500,3200)              
13.949                                     13.888

difference:                                    
-0.061 mag                               -0.005mag

Thus, if one is interested in doing 1% photometry, one either needs  
to use sky flats, or else one is restricted to a 2000x2000 pixel 
region centered at 1500,1500.  I've added the figures and numbers 
to my web page (http://www.lowell.edu/users/massey/obins/y4kcamred.html;
see the link to http://www.lowell.edu/users/massey/obins/y4kcamflats.html)

Does this mean all of my data obtained this semester so far is useless?  
(The placement of my fields have been pretty random, and I am trying to 
do time-series analysis of several hundred stars at the <1% level, 
differentially.)  I don't think so.  I actually wound up using a sky 
flat from a few nights earlier for the second test, so in principle I 
should be able to make a correction function.  Whether or not this will 
recover millimag precision, I dunno.

---phil