Re: Tale of a Tile

[Stupendous Man <richmond@stupendous.cis.rit.edu>]

> If I run with fixed sky tiles, then we will want to keep the tiles
> separate forever.  This means that a star measured on one edge of a tile
> will have to always be considered a different star from when it is
> measured on the other edge of a tile. Well, it's the same star, but users
> of the data will be urged to only look at the data from one tile at a
> time.

> This means that we don't want to have to look in the .fits header to sort
> out all the images from a tile.  Well, I don't think we do.

  Most users will get the data from a database; for them, the file
names are irrelevant.  A few very people -- like Tom -- will
view and analyze the data while it is still in its original
files.  If this name change makes their job easier, and doesn't
affect people downstream, then why not do it?

> This also means that we will want to put ra,dec into the .cal file names. 
> My plan is to have .cal files associated with a tile in the sky.  Else we
> lose the scatter improvement for the stars that appear in two tiles.

  Regardless of the file name, as long as one has in the database
a record of the original image from which a measurement was taken,
one can

              - pick out images which had the same center
                    (if desired)
              - analyze stellar measurements from these images
                    only

  I've written scripts to do exactly this.  

>  ... For the purpose of studying variable
> stars, the fixed position in a tile is better, and no worse in accuracy. 
> So why not measure that way?

  I agree, with one caveat.  I'd like to see the tiles overlap
by, oh, a half-degree or so, so that several hundred stars
appear at both left and right, or at top and bottom, of adjacent
tiles.  This gives you some idea of the size of the systematic
error in photometry which is hiding in the data.

  Perhaps one could make almost all runs in the non-overlapping
tile mode Tom has suggested -- which gives the maximum sky coverage 
for a single night -- and just once or twice a month (or season?),
one might choose a more tightly packed centering scheme so that
there would be overlap.  One might then derive the systematic
error from the few nights, which would be a very good check
on the instrumental response, if nothing else.  "Hey -- the
upper-left corner is a lot less sensitive than it was last
month .... oh, there's a bird's nest in the lens."

                            Michael