The new table of secondary standards (included) is generated from new global
photometric solution (solution #4). The difference between this solution and 
the previous one (#2) is the following:

1. New primary for field 90547 (the one that Sherry sent me) is used. 

2. North and South data are treated separately, i.e. different sets of nightly
   photometric constants a_J, b_J are obtained for each hemisphere. The previous
   solution generated a single set of constants for North and South data combined.
   Due to differences in photometric conditions between North and South that was
   increasing the residuals and making the overall solution worse. Now, there
   are two separate solutions and each one is tied to its own set of primary
   standards. Since there are common standards between North and South, the two
   solutions are thereby are tied together. Comparison of the calibrated 
   photometry between North and South reveals that the difference is at most 3%.

3. Another difference from the old solution is dropped nights. M. Skrutskie visually
   inspected the photometric residuals plots for all Northern and Southern nights 
   available to us and marked some of them as non-photometric, due to non-linear
   variations in sensitivity during the course of a night, or large residuals. 
   The list of dropped nights is available. The new solution is calculated with the
   data from these nights ignored.

4. Due to reduced size of the data set (North and South are treated separately) it 
   has become possible to increase the number of tracers per field. Now, the solutions
   for North and South use two tracers per field in addition to Persson standard.
   This leads to increased accuracy of the nightly parameters (the old solution 
   used only one tracer per cal. field).

5. As North and South solutions are tied to their own sets of primary standards, each
   hemisphere has its own table of secondary standards. We combine two tables by 
   identifying the same stars in common fields and assigning the variance weighted
   average to the resulting magnitude:

   m_NS = (m_N/sigma_N**2+m_S/sigma_S**2)/(1/sigma_N**2+1/sigma_S**2)

   where m_N and m_S are North and South average magnitudes for the star and sigma_N,S
   are root variance in the mean. This ties hemispheres together with complete
   account for photometric quality and repeatability in North and South.

6. Note that errors in magnitudes have dropped significantly (see Table). This is because
   we have quoted root sample variance before and root variance in the mean now.
   Variance in the mean is a factor of 1/n smaller than sample variance and is truly
   a measure of the errorbars on the listed magnitudes.

7. New table lists secondary standards whose calibrated magnitude is within the range
   [9.5, 14.5] and the toal number of observations more than 100 (if a field is common
   to North and South, the sum of North and South apparitions must be > 100). The new
   table is more extensive than the old one, because the major rejection factor in the
   old set was variance criterium (sigma < 0.05). Now that we use a root variance in the
   mean this is virtually a non-factor.

Notes:

  The new solution has been compared to the old one for matching secondaries. Results 
  show some differences (sometimes as large as 0.04 magnitudes). 


  The web page with all additional information and all plots is coming up soon.

--Sergei.