Minutes of the June 12, 2013 LBNE Sim/Reco meeting
Present:
Tom Junk, Eric Church, Zepeng Li, Jonathan Insler, Alan Hahn,
Matt Szydagis, Kevin Wood, Brett Viren
Apologies to those omitted
Matt and Eric have been working on using NEST to model the photon
production in LArSoft. Both Matt and Eric have made versions
of the simulation code that use NEST for the photons but not yet
propagate the electrons. Eric's involves an inheritance
of OpFastSimulation to hook in NEST's G4S1Light.
Matt has made a parallel development with the same functionality.
We still need to get the electron simulation hooked in to this,
as NEST generates them in an anticorrelated way. G4 calls the
optical and the electron modeling routines on each step, and we should
see which one is called first. At that point, the NEST model can
decide how many quanta are electrons and how many are photons,
save that in a list, so that when the other model needs a number
of quanta it gets NEST's prediction. Radiologicals skip the G4
step for speed, and will have to be NESTed too.
We should try to be flexible in the handling of propagation once the
quanta have been divided. For photons, we have three options --
full G4, library lookup, and a formula-based parameterization (see Zepeng's
talk).
Tyler has been working on CSU's cryostat geometry and has been working
with the CSU group to bring the simulation to reality. They have
been going down the path of building a photon library using Ben Jones's
tools developed for MicroBooNE. Issues involved here are that the
cryostat is a cylinder and may not accommodate the voxels as used by
those of us with more rectilinear geometries, and the fact that
the CSU cryostat does not have a TPC. Ben's routines input
TPC dimensions but do not make use of them. Tyler will talk with
Ben to see if they can be taken out, and the CSU people can take it out
of their private version for now (note -- it is always best to keep
versions from diverging).
Tyler has also been developing disambiguation algorithms. He has generalized
the channel intersection method to compute wire intersections and
give a list of all such intersections in the wrapped geometry. For simple
events with little activity on each time tick, disambiguation is easier
since corresponding U,V, and Z hits come in at compatible times. For
more complicated events with activity in more than one place in the same
time slice (cosmic overlay, complicated showers, or activity on the other
side of the APA), more work has to be done. Tyler is building lists of
U,V pairs and assigning Z hits to them.
Note -- hits can share others by view -- if a track's direction is more
closely aligned with the wires in one plane than with another, then
it will leave more hits in one plane than the other, and leave wider
hits in the plane with fewer hits. Eric suggests using the hit charge
and width as additional information for pairing hits together. We are thinking
about the best output -- could be just disambiguated hits which can be
clustered later, or spacepoints or clusters.
Zepeng has built a formula-based parameterization of light propagation
in the 10 kT far detector, and provided nice visualizations of it.
He has used a full G4 simulation of photon propagation to see what
the attenuation length is including Rayleigh scattering and absorption
using nominal LArSoft parameters. CPA and field cage reflections are
included in the G4 simulation with 20% reflectivity from stainless steel.
Zepeng added in the attenuation function along the bars from Ben,
which only goes up to 1m (limited by measurements), extrapolated to the
length of the LBNE FD acrylic bars. The efficiency falloff is quite
steep -- more than a factor of 10. Stan mentioned that the TPB coating
worsens the nice reflective properties of the surface and total internal
reflection is less efficient. A possible mitigation is to put SiPM's
halfway along the bars in along the midline of the APA frame. This
would double the number of channels.
Zepeng showed the photon response from 1.5 GeV muons -- they are
expected to go through three APA's and light is seen in roughly six.
Zepeng and Tom have asked Stuart Mufson for efficiency normalizations
for inclusion to see what the absolute yield is expected to be.
To do: flash-finding reconstruction and think about a good way to
incorporate photon detector information into the event display.
Jonathan Insler has been working on hit finding. Tyler noticed
about a week ago that there is a fraction of induction-plane hits
that have very little charge. Jonathan was able to reproduce this
effect. Jonthan and Tom suspect that it may be due to an interaction
of the zero-suppression and deconvolution processes. Deconvoltion
may assume that all small ringing and wiggles are present in order to
make sharp deconvoluted output. But if they are suppressed away, then
ringing appears in the deconvoluted output, producing spurious hits.
To test this hypothesis, Jonathan ran with zero suppression turned off.
Oddly, most of the low-charge induction-plane hits went away (though
there still are a few left), but more low-charge collection-plane hits
materialized.
Jonathan is working on a time-domain deconvolution -- this will allow
us to get away without unpacking the zero-suppressed data and use it
in place, and also allow us to limit the time extent of the deconvoluted
impact of a single bit of charge, reducing artifacts.
A tweak to the zero suppression algorithm may be needed -- add a few
time ticks (the number should be fcl-controllable) on either side of a
nonzero block.
A software trigger for 35T may have to be even more efficient -- approximate
hit finding without deconvolution may be needed in order to make the
trigger decisions fast enough.
There are minutes attached to this event.
Show them.