[d2n-analysis-talk] triggers on BB side (multi-hit TDC warnings)

[Brad Sawatzky]

On Thu, 05 Nov 2009, posik at jlab.org wrote:

> For run 1258, the T1,T5 and T6 seem to show good self timing peaks, T3
> seems similar in structure to what Dave sees, and the T2 Shows some double
> peak structure. Could this some how be a result of the shower cerenkov
> overlap? On the T2 there also seems to be a third peak to the right of the
> second peak. However looking at the production run 2086 this third peak in
> no longer present, and T1 for 2086 now seems to show two peaks.

A warning/reminder.  All the TDCs in our experiment were multi-hit.  You
need to be careful when you are working with multi-hit data:
ie. which hit is the one you care about?
    randoms, and electronic glitches like ringing and reflections can cause
    confusing/wrong results (incorrect and/or double counting, etc)

ROOT also does 'helpful' things based on context when you histogram a an
array variable.  For example, you might expect this command to histogram
'tdc_bar' with a cut on 'tdc_foo':
    T->Draw("DL.tdc_bar", "DL.tdc_foo-1000)<100")
What ROOT will do is this (in pseudo-code):
    for(i=0, 1, 2, <sizeof(DL.tdc_bar) ) {
      htmp->Fill("DL.tdc_bar[i]") if( "DL.tdc_foo[i]-1000)<100" )
    }
    htmp->Draw();
which might not be what you intended...

Analogous problems crop up when you histogram a multi-hit, integrate
over it, then apply a cut, integrate again and take a ratio of the two
to figure out an efficiency (for example).  Depending on what is in your
cut, ROOT might be building the histogram in two completely different
ways.

Bottom line:  Always include an explicit index when using any array
              variable.  You have to be particularly careful with 'DL'
              (ie. DecData) class variables, since they are /all/
              treated as multi-hit (ADCs too).

Bottom line 2: The default classes (THaCerenkov, THaShower, etc) only
               expose one hit to you, even if the underlying module is
               multi-hit.  This is a serious shortcoming when you have a
               high rate signal (ie. BB Cerenkov, BB Scint, in
               particular).  We will have to update the code to deal
               with this correctly -- a non-trivial bit of work.
               NOTE: For the BB Cerenkov TDCs, something like ~30% of
               the correlated hits not in the single hit the Cerenkov
               class propogates to the user.

-- Brad

-- 
Brad Sawatzky, PhD <brads at jlab.org>  -<>-  Jefferson Lab / Hall C / C111
  Ph: 757-269-5947 -<>- Pager: 757-584-5947 -<>- Fax: 757-269-7848
The most exciting phrase to hear in science, the one that heralds new
  discoveries, is not "Eureka!" but "That's funny..."   -- Isaac Asimov