Minutes of the February 13, 2013 Liquid Argon Simulation Meeting
Present: Tom Junk, Eric Church, Mike Kirby, Ben Jones
Tyler Alion, Mary Bishai, Norm Buchanan, Jonathan Insler,
Denver Whittington, Zepeng Li,
Apologies to those omitted
Tyler is nearly finished with the LBNE geometry -- he has checked
in a working GDML file with perl scripts that make its ingredients and it,
and the sorting of TPC's, wire planes, and wires is implemented. This
defines the numbering of the TPC's, planes, and wires.
Coordinates: The origin of coordinates is midway in the separator wall
between the two crystats, halfway up from the floor along the Y direction
(between the lower APA rows and upper APA rows), and Z=0 is defined to be
the edge of the active volume on the side the beam comes in on.
TPC zero is the lower one in the positive X cryostat closest to the origin.
X is the coordinate that increases first as you move to TPC 1, 2, 3,
then Y, then Z. The wire planes are sorted thus: U is plane 0, V is plane 1,
and the Collection plane (proposed to change it to kZ in the future,
still referred to as kW in the code) is plane 2 (this is true for all
TPC's regardless of which direction the field points). Wire sorting is listed
in Tyler's talk, LBNE DocDB 6813. Tyler has run the geometry through checks --
small overlaps at the 1E-15 level were found, presumably due to double precision
rounding. Setting the threshold to 1E-5 for checks makes all but a few
nonfatal overlaps go away. The service building and hill rotations need
updating (Arb8 rotations).
The APA frames now have the extra support beam in the middle, perpendicular to the
photon detector bars. They are simplified as two parallel metal bars instead
of metal tubes. The APA frames are solid steel but Bo gave Tyler the
steel thickness and Tyler will make them into tubes with liquid argon inside.
Tyler was wondering if the photon detector bars were twisted at the end for feeding
in to the SiPM's. Mike and Tom remember Stu saying that they are no longer twisted.
We will probably parameterize the response of the bars (Ben sent around measured
and predicted light response from MIT's lightguide R&D program. It's not quite
an exponential, due to the geometrical effects of how many internal reflections
light must make on its way to the SIPM.
CPA's are in their own volumes as subvolumes of the cryostats.
Tom has been able to push a muon through the simulation using yesterday evening's
geometry from Tyler. The logfile does not indicate problems, and an event rootfile
was produced. A naive attempt at using the event display crashes; Tom will
investigate. Tom switched off the noise simulation and included Huffman coding
to keep the memory usage in line.
Zepeng has been working to trim the size of the photon library. Taking
advantage of the symmetries -- all field cages are alike, and there are
three reflection symmetries within the field cages, gets the library size
down to 5 GB. As it currently stands, the photon library is loaded into
memory and thus it will not fit into our batch workers' memory.
Parameterizing far light and looking up near light in the library
may be a way to make the size manageable. Rayleigh scattering randomizes the
far light angles and so the response will likely be smoother. Cherenkov
light, which is expected to be 1/6 of the total, is not simulated well in the
library as the emission is assumed to be isotropic. Even if we do not parameterize
the far light and stick with the voxel/bar slab parameterization, the pairs
of production/detection pairs that are far away will be difficult to fill with
enough simulated photons and we may have to smooth it over anyway.
Matt showed a comparison of NEST's predictions of R vs. dE/dx, showing that
a power of -.85 intead of -1.0 in Birks' law works, without introducing
additional empirical factors that Icarus.
Incorporating NEST's electron/photon model into LArSoft: NEST now switches
between the Thomas-Imel mdel and Birks' law, depending on the particle
type and energy. Some work is involved in tracing the parentage of all particles
that ionize the argon so that an appropriate choice can be made. NEST uses
a similar diffusion parameterization model as LArSoft (electrons are drifted
parametrically instead of collision by collision with G4 which would take
too much CPU).
The Fermilab G4 team volunteers to work more closely with us, especially on
nuclear modeling but we would like to solicit their help on physics modeling
in tracking and ionization and physics lists. Matt has been working with
Mike Kelsey and Peter Gumplinger
Jonathan Insler reported studies of zero suppression and Huffman coding using
the MicroBooNE geometry and standard noise simulation, using single muon and
single electron events. Huffman coding gets a compression factor of 24-27.
Zero suppression with a cut of 10 ADC counts gets it down by a factor of 65 to 72.
MicroBooNE DAQ people will implement Huffman coding of zero-suppressed data.
Jonathan is collaborating with Herb and Andrzej on the noise model. Tom
mentioned that the parameterization of noise at the beginning of a job
is not very efficient for LBNE -- it stores about 5 Gbytes of random numbers,
in RAM, more than can fit in a batch job. This can easily be trimmed. It
has a nice feature in that the spectrum is explicitly specified and an FFT
gives the time representation of it.
Progress is being made on MicroBooNE's signal shaping modeling and also
the noise filter and signal shape deconvolution -- see the February 13
general LArSoft meeting slides at
https://indico.fnal.gov/conferenceDisplay.py?confId=6428
Jonathan volunteered to be the point man for making LBNE versions of these.
This is a good way to get involved in the reconstruction as well, as the
noise filter and deconvolution are the first steps in reconstruction.
March LBNE Collaboration meeting plans:
We will probably have, just like the Houston meeting, at least one parallel session
and plenary talks for simulation and reconstruction. But the spokes would
like a day-long pre-meeting for LBNE software. We would like to get people working
on software together to share ideas and define interfaces between groups.
Rather than duplicate talks to be presented at the collaboration meeting, we
envision at the moment a working workshop. We'll have an overview and a few talks,
and then breakout sessions on:
Beam Simulations
Interface with other groups: FastSim for physics sensitivity reach
Interface with ND and FD groups: flux files but also ntuples with particle
parentage so that cross sections can be reweighted and beamline
parameters can be studied. Other deliverables?
ND simulations -- Tom to contact Christopher Mauger and Kevin Yarritu on this
effort, and Alysia Marino
FD simulations and reconstruction
Physics modeling, charge/light modeling
geometry
digitization
disambiguation / (incl. event display. We hope to have a working
channel-level event display where the user uses his brain to understand
the mapping, and we should think of good ways to represent the data).
FD reco
Interface with fast simulation/physics group to predict sensitivity
There are minutes attached to this event.
Show them.