THE FOLLOWING ARE VERY GENERAL GUIDELINES FOR CHOOSING REFERENCE STARS
FOR PARALLAX TARGETS.  WE HAVE HAD MANY, MANY, DISCUSSIONS ABOUT
DETERMINING EXACTLY WHAT PROTOCOL CAN BE USED TO SELECT REFERENCE
STARS, AND THE BOTTOM LINE IS THAT THERE ARE NO HARD AND FAST RULES.
IN PARTICULAR, THE S/N > 100 RULE FOR A REFERENCE STAR IS SIMPLY TOO
RIGOROUS.  SOMETIMES YOU JUST HAVE TO GO FAINTER, AND SOMETIMES THAT
IS TOO FAINT FOR A GOOD REFERENCE STAR...

INSTEAD, JUST KEEP IN MIND, 

     "I WILL DO THE BEST I CAN WITH WHAT NATURE HAS PROVIDED."

----------------------
WHEN TO REDUCE
----------------------

Generally, it is not worth reducing a field for parallax unless you
have at least 2.0 years of coverage, and at least 20 frames in the
evening plus 20 frames in the morning.

----------------------
BEFORE FIRST REDUCTION
----------------------

1. Ideally, you first crosscheck the parallax setup frame against the
   frame(s) on which you will do photometry.  If any star near an edge
   has fallen off the VRI frames, you can't use it for a reference
   star because you won't have VRI for refraction and the correction
   to absolute parallax.  If you don't have the VRI photometry frames,
   yet, keep the stars along the edge as reference stars.

2. To select reference stars, choose a single frame in the parallax
   filter that has seeing better than 1.5" and has ellipticity better
   than about 0.2.

3. Each possible reference star should be checked to make sure that it
   is not a close double star or a galaxy.  Either type of source will
   result in high residuals.

4. You want a balanced reference field.  So, choose a reference star
   configuration that surrounds the target star (if possible).

5. You should have at least 5 reference stars, but not more than 15.
   Tests have shown that more than 15 reference stars typically makes
   the residuals worse, not better.  (We don't know why this is, but
   that's the way it seems to work.)

6. Usually, the peak count must be > 100 for a potential reference
   star.  This does NOT mean that anything above 100 counts is useful.
   You really want to stay above about 1000 counts, but can try going
   lower if you don't have 5 reference stars.  

   However, sometimes the reference field is so faint that you do need
   to use reference stars with counts < 100.  Tests indicate that as
   long as there are enough reference stars and enough frames, the
   final values for parallax and error do not change by more than 1
   sigma.

----------------------
AFTER FIRST REDUCTION
----------------------

... more details to come here, as we get a better feeling for what
residuals seen in parallax and proper motion are too high to keep,
using the new pipeline ...

7. parallax residual cutoff = ???

8. proper motion residual cutoff = ???

9. high residuals on frames = if one star is bad in 50% of frames,
   discard it.

-------------------------------------
EXAMPLES OF DISCARDED REFERENCE STARS
-------------------------------------

DEN1048  #7     faint
         #10    faint

GJ1025   #4	highest residuals in many frames

GJ1068   #2	faint
         #4	phot pi > 2 mas
         #10	faint

LHS0158  #4     highest residuals in many frames
	 #8     bad column

LHS0193  #3     trig pi = 27 mas !

LHS0288  #3	highest residuals in many frames