Post on 14-Jan-2016
BigCal Reconstruction and Elastic
Event Selection for GEp-III
Andrew Puckett, MITon behalf of the GEp-III Collaboration
Introduction• Experiment E04-108 will measure the proton form factor ratio
GE/G
M to Q2 of 8.5 GeV2 using the polarization transfer method.
• Scattered protons are detected in the HMS using parts of the
standard detector package—drift chambers and S1 scintillators.
New scintillator S0 forms custom trigger.
• Transferred polarization is measured using a new FPP built by the
collaboration (Dubna).
• BigCal, a large solid-angle electromagnetic calorimeter, detects
the electron in coincidence with the proton and is part of the
trigger.
• Timing and kinematic correlations between BigCal and HMS are
used to reject inelastic backgrounds
HMS Detector Package for GEp
Scintillators S1 and S0 (new):
Trigger and timing
HMS Drift Chambers:
Track protons
FPP Drift Chambers:
Track scattered protons
CH2 Analyzer
HMS Shower Counter
BigCal—Detect Scattered Electron
•1744 lead-glass
blocks equipped
with PMTs
•4” Al absorber in
front reduces
radiation damage
•Light source--•Lucite plate
illuminated by
LED via fiber
Floor Layout of BigCal
HMS Trigger
• Nominal Settings:
1. Require PMT at both ends of paddle to fire
2. Require S1X and S1Y for “S1” trigger
3. Require S1 and S0 for HMS trigger
4. Two different trigger types for HMS at T.S.—one for each paddle of
S0
Different logic was used at different times to check
efficiency Non-standard triggering affects TOF calibration
BigCal Trigger
• Apply high threshold to the
analog sum of 64 PMT signals.
• Summed groups overlap
vertically, improving efficiency
• To get best efficiency for this
trigger, phototube gains must
be fairly well-matched—
calibrate HV using elastic ep.
Coincidence Trigger
• Trigger signals are timed so that BigCal trigger arrives first, about 15-20 ns before HMS trigger
• This way, the HMS scintillators determine the timing of all ADC gates and TDC stops(or starts) for
true coin. events.
• Width of coincidence timing window is 50 ns.
Trigger Rates
Rates in this table in kHz
Trigger Rates, cont.
• Accidental coincidence rate estimate for kin. 5:
• 11.6 kHz HMS2 triggers (elastic paddle of S0)
• 621 kHz BigCal triggers
• True elastic rate < 1 kHz << HMS/BigCal rate
• Poisson Statistics—probability of random BigCal trigger
given HMS trigger:
BigCal Reconstruction
Three main tasks for GEp:
1. Energy reconstruction
2. Position reconstruction
3. Timing
Energy calibration can
be updated continuously
for elastic ep—straight-
forward linear system.
Position requires shower
shape determination
Timing—offsets and
walk corrections
Cluster Finding Strategy
1. Find largest maximum
2. Build a cluster by adding nearest neighbors with hits
3. Work our way outward—allow clusters to expand freely in any direction
4. “Zero” hits in the current cluster
5. Repeat 1-4 with remaining hits until no more “maxima” are found
Energy Reconstruction
•Electron energy is known to within ~1% from HMS
momentum/elastic kinematics
•Chi-squared minimization gives a system of linear equations in
the calibration constants—determine as often as needed for GEp.
•Have to solve system of 1,744 equations!
BigCal Position Reconstruction
•Observable quantities are shower
“moments”: energy-weighted mean
block positions
•Moments vary with distance of
electron impact point from center of
max. block.
Shower Shape Determination•Distance from block
center varies non-
linearly with measured
moment
•Fit “S” correction to
the distribution of
impact point vs.
cluster moment.
•Tracks incident at
large angles have
distorted shower shape
Position Resolution
•Using BigCal monte-carlo
developed at Protvino, coordinate
resolution betwen 4 mm and 1 cm
is demonstrated
•Determination of true shower
shape considerably more
complicated
•This example has 4” absorber,
~1.2 GeV electrons
BigCal Timing
• Blocks are timed in groups of 8:
32x56/8 = 224 TDC channels
• The major correction to the
measured time is an offset for the
slightly (or very) different cable
lengths.
• There is also a significant pulse-
height dependence to the
measured time that can be
corrected for.
• Timing information is also
available from TDCs of the sums
of 64 used to form the trigger.
Cable Length Offset
Hit times relative to BigCal trigger
•TDC hits come in at a
nearly constant time
relative to the trigger
•Find peak position in
TDC spectrum to
determine offset
Walk Correction
•Hit time has a
significant pulse-height
dependence
•Determine for each
group of 8, do simple fit
•Apply correction to hit
times
Sample time-walk profiles for groups of 8
Cluster Timing• Throw away TDC hits outside a
window of about 150 ns ( 75 ns of
BigCal trigger time). Such hits won't
have corresponding ADC hits within
the gate.
• Within clusters, find all TDC hits in
corresponding groups of 8. If multiple
hits, take the hit which best agrees
with the maximum.
• Compute energy-weighted mean and
rms times.
• Timing resolution ~3 ns
Elastic Event Selection
• HMS measures proton momentum and angles.
• With BPM and raster info, we can correct reconstructed
target quantities to determine IP
• Correct BigCal angles using the ray from the HMS vertex to
the reconstructed BigCal position
• In the case of multiple clusters, use HMS to pick the best
cluster assuming elastic kinematics:
HMS momentum-angle correlation
• We can select elastic events by
looking at vs in the HMS by
itself.
• Some kinematics still have
substantial inelastic backgrounds
under elastic peak.
• To put FPP in HMS hut:
– No PID capability (no
gas/aerogel Cerenkov)
– Limited timing resolution
(no S2)
• Need BigCal to clean things up:
– See effect of various BigCal
cuts in figure-->
HMS momentum-angle correlation
HMS-BigCal Correlation
Remaining Tasks
• Use survey data to fine-tune geometry definition
• Check BPM/raster corrections
• Optimize cluster finding parameters/improve the code
• Improve/optimize parameter database for large-scale
analysis
• Determine shower shape parameters from the data
• Write 0 reconstruction code for multi-cluster events
Conclusion
• BigCal is successfully serving its
purpose as electron detector for
GEp-III
• Some work remains to be done on
analysis code
(clustering/pions/shower shape/etc)
but things looking good so far
• Clean elastic event selection for high
Q2 GEp-III and low-ε GEp-2g