Jan Conrad, 020830

Updates:

- Some of the monitoring files have been produced doubly.
  This leads to a correction of the effective livetime, see below.
  (020916)

  Also the runtuple has been updated:  run_99.hb  (020916)

Update history:

 - The ntuple with run information can be get  here  (020911).
 

The effective livetime of AMANDA  is (as of 16/09/2002) :
 
 

                                                                                                    222.63 days
 

here BAD runs  as defined in the OM/RUN selection are removed.
 
 
 
 

If you want to exclude all BAD runs and runs  with run number < 175 then the effective livetime of AMANDA in 1999 is:
 
 
 

                                                                                                     192.27 days.
 

The exclusion of runs < 175  might be motivated from this  plot, .
 

 For comparison the total wall clock time of data taking   (using this run selection)
 was 232 days (day 83 -315), which is approx. 20 %  higher than the effective livetime.
 

The five steps to get this number :

 1.  an exponential fit to the delta t distributions (scale  < 50 ms )  of the monitoring data
      has been performed. An example of the delta t distributions can be seen here

      The true rate is then given as:
 

                                                  true_rate = 1/ (slope_exponential)                                     (1)
 

 2.  raw rates have been calculated from the run-statistics files, by taking:

                                                   raw_rate =  N_events /  file_duration.                                   (2)
 
 

3.  The  effective livetime of each file is calculated by:
 

                                                  Livetime_eff (file) = file_duration * correction_factor       (3)

   where
                                                   correction_factor = raw_rate/ true_rate                               (4)
 

4. The effective livetime of AMANDA for the whole year is consequently:
 

                                     Livetime_eff(1999)  = sum_(files |  run selection) ( Livetime_eff(file))    (5)
 

     where the sum is taken over all files within the run selecion.
 
 

 5.   File length have been checked . There are   13 files  with abnormally large duration. They fall all
       in the period < run 175.  If the file length is due to GPS problem in a small number of events, the method used
       here will compensate for it, thus given a similar time if they are included or excluded. Indeed, excluding these files
       gives a lifetime of 224.15 days, which is strongly supporting the hypothesis that it is a problem of few event times.

      Hereis a plot of file duration  ( units are seconds) distributions, and here is a plot of  accumulative  effective file duration
 

The exponential fit to the 22698 monitoring files has been done using this   root macro

The lifetime calculation (taking the runstatistics file and the output of the root macro as input ) has been performed
using   this perl script  (thought of as templates, have to be modified for your needs).