[d2n-analysis-talk] Spike at E/p=0 [SOLVED, but begs a new question]

[Brad Sawatzky]

On Thu, 16 Dec 2010, David Flay wrote:

> On Thu, Dec 16, 2010 at 11:18 AM, Brad Sawatzky <brads at jlab.org> wrote:
> 
> > There are four ways E/p == 0:
> >  1) the code assigns E/p = 0 as a default value,
> >  2) the code assigns E/p = 0 as an error/bailout,
> >  3) the quantity E really is 0, and/or
> >  4) the quantity p is huge (ie. kBig).
> >
> > We need to know which of these cases is contributing to the
> > non-physical E/p==0 number, and whether or not that makes sense.  If
> > some are due to trigger or hardware problems, then we need to make
> > sure our efficiency corrections treat it properly.
> >
> > The single adc histograms for a single PMT aren't really sufficient
> > since the majority of the zeros there just mean that that particular
> > cell didn't interact with the particle (which is quite reasonable).
> > What I would like to see verified is that the raw sum of the energy
> > in the pion rejector layers equals the (summed) pedestal value.
> > That is, I want to verify that getting E=0 out of the "high-level"
> > code really means that we had negligible energy deposited in the
> > cells at the lowest level.  That would be case '3' in the list.  If
> > that's what is happening, then it is strange:
> 
> Do you mean cutting on L.prl1.e==0 and seeing what shows up in
> L.prl1.asum_p (same for prl2)? I don't have a variable for the raw
> sum.  In which case, the events at zero for L.prl1.asum_p and
> L.prl2.asum_p do show up at zero (blue curve in attached *.pdf).

Do whatever you need to do to explain which of the options I listed
above are happening.  I have some suggestions below, but they are not
the only way to investigate this issue.  You know that E/p=0 should
never be zero for good particles.  The software is reporting a condition
that can't be true for the physics you are investigating with your
system.  Something in that system (LHRS + DAQ + software) is reporting
garbage.  If it is happening 0.1% of the time, then we don't really care
(it's a little disconcerting, but a rare hardware glitch, timing issue,
etc can do that).  If it is happening 10% of the time, it's a real
problem.  Now you need to think about how to divide the problem in to
pieces to figure out where the non-physical zero is coming from.

The first thing I would do is evaluate how important the E/p=0 problem
is.  It would be nice to know how often that condition occurs as a
fraction of:
  - total physics triggers, and
  - physics triggers with a reasonable track (ie. so momentum should make
    sense at least)
Some of your plots seem to indicate E/p is happening for more than 10%
of physics triggers that have a track.  Is that really the case?
Are those E/p==0 values correlated with cosmics?  (ie. the frequency of
the problem will get larger at higher LHRS p0 setting, beta will be
negative, the track will probably not reconstruct to the target).

Next, or as part of the above replay, I would construct my own prl1 and
prl2 sums by adding all of the raw block ADC data together:
  MySum = L.prl1.a[0]+...+L.prl1.a[n] + L.prl2.a[0]+...+L.prl2.a[n]
I would then histogram that quantity vs. the total energy value that is
going into the E/p ratio.  If the software is working correctly, then
you should see a nice linear correlation between MySum and E, all the
way down to the non-zero pedestal in MySum and E=0.  If there is a lot
of scatter, or there is a band where E=0 but MySum != pedestal, then
something is wrong.

You can also generate some histos that will look at the momentum side of
E/p.  Is E/p==0 because p==kBig?  What are the conditions that are
leading to p=kBig, and is that a reasonable thing for the software to be
doing under those conditions?

-- Brad

-- 
Brad Sawatzky, PhD <brads at jlab.org>  -<>-  Jefferson Lab / Hall C / C111
Ph: 757-269-5947  -<>-  Fax: 757-269-5235  -<>- Pager: brads-page at jlab.org
The most exciting phrase to hear in science, the one that heralds new
  discoveries, is not "Eureka!" but "That's funny..."   -- Isaac Asimov