Stale data monitor
updated Tue Nov 2 09:09:24 EST 2004

  • Goal : detect stale data in a EEMC crate
  • Method : calculate chi2 between events with the same token
  • Example of implementation:
    * store in memory last event for every token (max=4098)
    * compare two events with identical tokens, replace the old event afterwards
    * include in chi2 only channels which changed be less the 10 ADC counts (one can say it uses only pedestal values )
    * average chi2/DOF over e.g. 1 minute interval is a reasonable measure of none-staled data.
    * implemented for 6 crates of ETOW

  • plots illustrating the above algo

    1. total chi2 vs. event ID. In the run 5135068 only 128 tokens were in use, so after ~200 events there is one event in memory for every possible token and total chi2 stabilizes. The plot below covers just 1 minute of data taking.
    2. number of ADC channels per event used to calculate chi2 ( i.e. with |delta ADC| <10 ) is close to 95%.
    3. chi2/DOF remains on the level of 2.5 as long as pedestals fluctuate. Note sqrt(2.5)= 1.6 ADC, which is close to sigma of pedestal of about 1 ADC.

    Code used for this example is in CVS StRoot/StEEmcPool/muEztStaleData. Uses new muDst with Ezt-branches.