Number of photons per sq.cm. per sec in V-band from a V=7.6 star is

                     10^6 photons        -0.4*7.6
     N(star)  =   ---------------  *   10
                    sec * cm^2


                              photons
              =      912    --------------
                              sec * cm^2
 

   Number of photons collected per second in V-band is

                912 photons 
   N(coll)  =  -------------  *  pi * (180 cm)^2
                sec * cm^2

 
            ~    10^8   photons / sec


   In order to see a feature of depth 10^(-6), one must exceed
the Poisson noise in the starlight.  

                                       1
       starlight noise  =   ------------------------   
                            sqrt(number of photons)


   Therefore, the number of photons collected must be


         N(coll)  >   (starlight noise)^(-1/2)

                  >   [ 10^(-4) ]^(-1/2)

                  >   10^(8)


   How long should this take?  

                             10^(8) photons
          exptime   =      -----------------------
                              10^(8) photons/sec 


                    =     1 sec


   How long _did_ they actually expose?

               Three nights, several hours per night


   Why the difference?


               They broke up the light into spectra, instead of adding
               up all the light within a broadband filter.

               V-band filter bandpass width approx 900 Angstroms

               spectroscopy "bandpass" width approx  0.5 Angstroms