We cannot afford losing any triggered data in an uncounted manner. For absolute x-section measurements with GRETINA efficiency is crucial.

For S800-GRETINA data with GRETINA self-trigger only we find S800 events in Global.dat without corresponding GRETINA event.

In a low rate file, the following event counts were found:

     11,886 coincident events (s800 and gretina present with correlated timestamp)
     719    s800 only events
     15     gretina only events (now fixed in S800 daq)
     
So, ~6% of the decomposed data is missing.

Also, missing gretina data is spread throughout the file, not concentrated at the beginning or end.

• 5/17/2012: Discussed losses with Mario. Code review and version check to follow.
• 5/24/2012: (kathrin) checked runs 694 (old decomp) and 697 (new decomp) taken on 05/22

  coincidences = s800 and gretina GGEB header with time stamps within 1000 ticks
  s800 singles = missing gretina GGEB header and event
  gretina singles = no s800 in this event
  decomp fails = error in decomposition, "pad" of gretina struct is != 0 (I get errors 3 and 5 see below)
   
  Run 694:
      172171 coincidences (94.4 %) 
      10207 S800 singles (5.6 %) 
      18 Gret singles (0.0 %) 
      0 Decomp fails (0.0 %) 
   
  Run 697:
      97888 coincidences (91.2 %) 
      9477 S800 singles (8.8 %) 
      0 Gret singles (0.0 %) 
      81 Decomp fails (0.1 %) 
      
  decomp fails are included in the coincidences (since they have a valid time stamp), as error I get 3 and 5.
  in both cases the event is completely empty (everything is zero) except for the crystal_id.
  out of the 81 failures I have 54 error 3 (no net charge segments) and 27 times error 5 (bad chi^2)
  

this test has to be repeated, to me it is completely unclear, why run 694 has again S800 problems (these were thought to be solved before)