Thanks to Jeff Munn, Dan Vanden Berk, and Brian Yanny we have
made it up through the end of complete_tile, using the operational
database.  This is a major step forward, though I would not claim we
have achieved a production-ready system; we still need to do many steps
by hand.  We have a week left to run the outputs through the plate 
design code, and Aronne Merrelli will be here at the end of the week to 
make sure this happens.

	The results of target selection and tiling are available on
Jen Adelman's new web page:

http://www-sdss.fnal.gov:8000/~sdssdp/target/target.html

Here you will find plots and tables of target selection flags and
tile-by-tile plots of objects placed.

Here are some notes on this target selection run which are of interest
to the scientists on the project (these are operational observations only -
there are email messages from individual working groups who have analyzed
the successfulness of various algorithms):

TARGET SELECTION FLAGS:

   The galaxies module flagged *way* too many BRGs.  This necessitated
cutting the number of normal galaxies placed on each tile by a factor of
two to get the galaxy density down close to 100 per square degree.
See message on culling (http://www.astro.princeton.edu:81/sdss-target/msg.85.html)

   Other culled objects included tiled objects in bad columns, ``galaxies"
which had stellar colors as determined by Michael Strauss, qso candidates
in fields with presumably bad calibrations.  The culling step is not
part of the target selection production system, so these need to be
eliminated some other way in the future.

   There are many more ROSAT objects flagged than we anticipated.  When we
placed these objects as flagged, we got about 30 ROSAT fibers per plate.  The
code expected about 3 per plate.  Since ROSAT objects were expected to be
rare, they are placed on plates immediately after tiled objects, using as
many fibers as possible to get all of the ROSAT sources.  We believe that
two factors contribute to the over-abundance of these targets: (1) the
ROSAT catalog is denser than we expected, possibly going to fainter limits
than we anticipated, and (2) the radius an object is flagged from a ROSAT
source coupled with the ROSAT_D selection flag (which is applied to galacitic
stars within the error box) causes about 3 ROSAT fibers to be assigned per
ROSAT catalog source, the majority of which are very low priority.
   In this run, we added code in complete_tile to not target ROSAT_D sources
at all, and in addition to not target every other ROSAT source.  This way,
we reduced the number of ROSAT sources to about 5 per plate.  Someone needs
to fix this for the next tiling run.

   There are somewhat too many QSOs.  We note that many of the FIRST
quasars are in regions of the color-color plots where we expect to find
galaxies.

   For this run only, we tiled brown dwarfs to allow the stars working
group to check the target selection of this rare type.

   The QA plots for target selection are not adequate.  In particular
There only exist plots for quasars and galaxies (and even those are not
adequate).  Brian added tables with numbers of targets per field sorted
by type to partially alleviate this problem (these are also on Jen's
web page), but more work needs to be done.

TILING

   There continue to be many problems with the placement of object fibers
within each tile.  This run used 85% completeness, equatorial (so the
plates would not move in declination), and a maximum of 575 tiled fibers
per plate.  The actual number of unreserved fibers per plate in this run
is 579, so we purposefully reserved four fibers per plate for ``as available"
targets.  As it turned out, all of those were used for ROSAT sources, so
many tiles did not have any stars, serendipity, or QA fibers placed.

   We would like to bring to the attention of the collaboration that there
are fewer science fibers available per tile since we are putting down
50 sky fibers.  The large number of sky fibers is considered necessary
for calibration purposes, but be aware that this reduces the number of
galaxies per square degree for which tiling currently works.  (If we use
the same number of objects per square degree with fewer fibers, the plates
get pushed closer together).

   Currently when we run tiling we have to try a bunch of different input
parameters (completeness, sky density, fibers/plate) to try to get a
reasonable result.  We then evaluate the tiling outputs to see how good
the coverage looks and what fraction of the tiles look like they pass
the minimum fiber reach of the science fiber bundles.  The code seems to
favor sections of sky on which no objects are placed, especially verticle
strips of untiled objects and tiling none of the objects in the overlaps
between two tiles.  So, the untiled objects are in no sense evenly
distributed.  In general, the higher the efficiency, the fewer sections
of sky with untiled objects, but the larger the overlaps between plates.
Larger overlaps lead to more plates with fibers bunched all on one side,
which cannot be plugged.  Balancing these two factors led us to the 85%
efficiency choice, though different runs would probably be different since
there is a significant randomness to what effiency happens to look best.

  We produced 28 tiles, of which 11 used all 575 fibers allocated to
tiled objects.  There were an average of 21+4=25 fibers per plate
left for ``as available" targets.  We note that the plates with all
fibers used were tiles 30-40 inclusive.  We do not know why the full
plates would be in sequence.

COMPLETE_TILE

   As noted above, there is special-case code in this section to deal with 
peculiarities in this tiling run, especially for ROSAT and BROWN DWARF
candidates.  In order to give some fibers to stars and serendipity, we
changed the allocations of stars/serendipity/QA from 32%/18%/50% to
60%/30%/10%.  To stop the algorithm from assigning random quasars and
galaxies which were culled as QA targets (which were not actually assigned
to other tiles), we added a check to make sure the QA object was also
assigned elsewhere.  Even with this reallocation, there are many plates
on which no stars, and especially no serendipity objects, are assigned.

  We are now running with the complete Tycho catalog to find light trap
stars.  The magnitude limit for light traps is also brighter than the
previous tiling run.

  the output plPlugMap files now contain the target flags to make checking
the results easier.

					- Dan, Brian, and Heidi