Here's a cookbook for the latest method using the ucsclris software with
pre-selected guide star and the pickoff mirror:

AFTERNOON:

1. Create the Box Coordinate Files:
    Write a text file for each mask, containing (one line per star):
	x  y  (Optional ID)
Take the x,y CCD coordinates from the "map" file created by mapmask.

2. Confirm the Coordinates:
    Take an image through each slitmask (you may use internal illumination,
eg, Ne lamp and V filter, 2s each; or the dome).  Run xbox to make sure
that all boxes are found (it doesn't matter where it finds a "star").
Note the delta values for each box: if they're constant and larger than
about 3--4 pixels, you may want to epar "xoff,yoff" in mboxfind.
(You may enter "I" when the residual plot is displayed to interrupt the 
program.)

3. Set up the STARLIST for the OA:
    Write a "starlist" file containing the Guide Star RA,Dec for each mask.
Note that the line contained in the header of the map file is close to
the correct format:  strip off the first 4 characters and replace the :'s
in the coordinates with spaces.  This starlist file is sent to the OA.

Also, note the PA's for each mask (slightly updated for precession), and the
TV pixel coordinates for the guide star.



FOR EACH MASK:

1a.  Give the OA the Guide Star Coordinates as identified in the starlist.
NB: Tell the OA these go at the PICKOFF Pointing Origin!  Tell the OA the
PA on the sky (updated for precession, see #4 above).

1b.  Insert the slitmask and move to the imaging mirror.  The OA should switch
to the slit-viewing camera.

2.  Identify the Guide Star in the slit-viewing camera.  Tell the OA to
place the star at the TV pixel coordinates in the "map" file (see 4 above).

3.  Take a direct image through the mask (30s is typically good).  Stars
should be seen in the boxes. (* if not, see Note_1 below)
Analyze the image:

	uc> xbox lris0023 box_coord_file

(if problems, see Note_2 below)  You should be able to type "q" at each plot.
Confirm the offsets; if less than about 0.03 degree in PA and 0.1" in RA/Dec,
don't bother --- you're there.  (Also, see Note_3)

4.  Repeat the last step as needed.  When aligned, switch back to grating and
go!


TROUBLESHOOTING:

NOTE_1:  If stars are not seen in the boxes, something is wrong.  Immediately
pull out the mask and take a direct image.  Display the direct image and
the mask image. Identify an alignment star and the box it should fall in,
and use "mshift" to move it there:

	uc> mshift (xstar) (ystar) xref=(xbox) yref=(ybox) pa=(PA)

(You should also "blink" the two images to make sure that the PA is close.)
Then re-insert the mask and return to step 3.  NOTE: if the PA error is
too large (one star will be in a box, but the others won't) see Note_4 below.

NOTE_2:  Things to watch for...
"Cannot find box" and crash!  Oh oh, this shouldn't happen any more if you've
followed the procedure above.  Here's what to do.  Display the image and
measure the coordinates of a couple of boxes.  If there's a constant offset
between the boxes and what's listed in the "box coordinate file", then epar
mboxfind to change the offsets (don't forget to return them to zero when
appropriate).  Note that _some_ offset may already be applied, as listed by
xbox -- be sure to _exclude_ this when changing xoff,yoff.
If only one box is failing, you can measure its coordinates in the image
AND SUBTRACT the offsets listed by xbox; incorporate these new coordinates
in the box coordinate file.  If the one box that fails is non-crucial,
you can simply comment it out (#) in the box file. Then rerun xbox.

There is no box (only noise) seen in the profiles.  This is the same case as
above, but the program doesn't crash.  Same options as above.

The vertical line marking the star center isn't where you think the star 
center should be:  place the x-cursor where you want it and type "f".

The sky isn't set right: place the y-cursor where you want it and type
"<space>".  The centering is not very sensitive to this.



NOTE_3:  If the stars are off at the very edge of the box, you may want to
have the OA apply the translation without the PA change; then take another 
image.  This is particularly true for crowded fields, where misidentifications
can really lead you astray.  In any event, the PA offset may not be very
reliable if the stars are mostly occulted.



NOTE_4:  The PA error is so large that even though one star is in the box,
other ones aren't.  This is a serious problem, and one that should not happen!
If it does, though, here is the fix:
1. remove the mask and take a direct image.
2. measure the locations of two stars (one at top, one at bottom of image).
3. measure the locations of the corresponding boxes.
4. write a text file ("pairs file") that lists (one star per line):
	xbox ybox xstar ystar
5. run xbox:
	xbox lris0027  pairs=pairs_file
6. Since you've hit serious problems, take another direct image and
blink it against the previous mask image to confirm the stars will be
in the boxes.
7. reinsert mask and continue at Step 3.
