• Main INDEX
  • Monthly INDEX
  • PREV
  • NEXT

    User name R. Michaels

    Log entry time 00:20:54 on June10,2005

    Entry number 145697

    Followups:

    keyword=deadtimes ok

    Bodo did a number of tests, and also inspected cables and looked
    for double pulsing. No problem with the hardware. From his tests
    its clear the "high deadtimes" only occur if there's a combination
    of triggers like (T1,T3) or (T3,T6) with prescales=1 while all other
    prescales are high. I think those are the only problematic combinations.

    The reason is understood:
    In the online calculation of deadtime, I compute a global number
    which takes into account correlations, albeit imperfectly.
    For example I assume all T5 are also T1 and T3, and that
    all T7 are T5. However, I neglect the correlation of T1 and T3
    that arises from pulser (except via T5) so if prescale for
    T5 is high this causes an error. For cosmics, the rate
    is ~10 Hz and the EDTM pulser is 2 Hz so this is a 20% apparent deadtime.
    Could fix it by subtracting off EDTM pulser, I suppose
    Similar problem for T3 and T6 which tend to be
    correlated for cosmics. However, here since we don't know the
    correlation coefficient of T3,T6 in general, datamon (and halog end-run)
    will always have this small mistake. I think it can't be fixed.

    Under beam conditions (~kHz) and with typical prescales,
    these effects are negligible.

    But.... it is *NOT* a real problem anyway because the deadtime is
    unambiguously determined for each trigger ... ratio of accepted
    trigger to input. Easy to do offline with "Podd". When done
    this way, the deadtimes of each trigger should be about the same.
    I haven't checked this yet... maybe will look tonight while on shift.

    It would be great if we could compute deadtime for each trigger
    online but we don't know how many triggers of each type are
    accepted by CODA -- not until we analyze a data file.