VictorGenty_FinalPresentation

VictorGenty_FinalPresentation
Electronics Testing, LArSoft Analysis,
and Data Acquisition for MicroBooNE
Victor Genty
Nevis Labs
August 1, 2013
Genty (Nevis)
REU Presentations
August 1, 2013
1 / 31
Outline
Genty (Nevis)
1
Mini & Micro - BooNE
2
Low Energy Excess
3
LArSoft Analysis
4
PMT Gain Study
5
Splitter Reflection
6
PMT Data Acquisition
REU Presentations
August 1, 2013
2 / 31
MiniBooNE
Studied:
νµ → νe oscillations, both modes
With:
Cerenkov detector, 950,000 liters of
mineral oil, 1520 phototubes in
12-meter diameter sphere
Found:
Observed data above 475 MeV are
consistent with expected
background
A low energy excess below this
energy
Genty (Nevis)
REU Presentations
August 1, 2013
3 / 31
Variety of interpretations by
many beyond the Standard
Model physics including...
3+N Sterile Neutrinos
...but could be misidentified νµ
→ can not distinguish e − and γ
signal
0.6
0.2
0.3
Data - expected background
0.2
Best Fit
2
2
sin 2θ=0.004, Δ m =1.0eV
2
2
sin 2θ=0.03, Δm =0.3eV
0.1
2
2
0.0
-0.1
MicroBooNE detector proposed
to study even lower ν energy
ν e & ν e from µ +/+/ν e & ν e from K
0
ν e & ν e from K
π0 misid
Δ → Nγ
dirt
other
Constr. Syst. Error
Best Fit (E>475MeV)
Fit Region
0.4
Events/MeV
Excess events in 200 - 475 MeV
neutrino energy region found by
MiniBooNE.
Events/MeV
Low Energy Excess
0.2
0.4
0.6
0.8
1.0
1.2
1.4 1.5 1.6 3.0
QE
Eν (GeV)
A. A. Aguilar-Arevalo et al. (MiniBooNE Collaboration), “Event Excess in
the MiniBooNE Search for ν̄µ → ν̄e Oscillations”, Phys. Rev. Lett. 105,
181801 (2010)
Genty (Nevis)
REU Presentations
August 1, 2013
4 / 31
MicroBooNE - Detector
Specifications
170 ton liquid argon cryostat
Time Projection Chamber (TPC) with 3
wireplanes
32-40, 8-inch photomultiplier tubes
Will study
LAr
νe /ν̄e appearance
Genty (Nevis)
REU Presentations
August 1, 2013
5 / 31
LArSoft - Detector Simulation
LArSoft is a complete set of
simulation, reconstruction, and
analysis tools for liquid argon
detectors
Whole detector simulated
by GEANT4 (LArG4)
Neutrino beams simulated
by GENIE, all other
particles possible
Reconstruction chain
developed
Event display for three
wireplane, can investigate
reconstructed parameters
against truth...
→
Genty (Nevis)
REU Presentations
August 1, 2013
6 / 31
LArSoft - Event Reconstruction
Reconstructing neutrino interactions inside MicroBooNE
Clustering
Hits are signal vs time information
from a calibrated Wire object and
looks for peaks that indicate real
energy deposition occurred
Clustering algorithms identify
reconstructed wire hits which are
correlated both spatially and
temporally
DBSCAN and Fuzzy Clustering
are two such algorithms
Raw Data
Wires
Calibrated Data
Hits
Clusters
Energy
Total visible energy deposited on
TPC from e − showers
Genty (Nevis)
REU Presentations
2D/3D Tracks
August 1, 2013
7 / 31
LArSoft - Cluster Studies
1
Generate νe events filter for
1e − + 1p final states,
simple event topology
2
I wrote a LArSoft module,
MCHitter, to calculate
purity and efficiency of
reconstructed clusters
3
Compare DBSCAN,
FuzzyCluster
⇒
Efficiency
Purity
Measures
Measures
How much of a cluster is composed
of a each true particle
How many of all hits the particle
generated are in a specific cluster
If less than 1: clustering algorithm
could not distinguish true particle
hits from one another
If less than 1: algorithm failed to
group the hits created by the
particle into a single cluster
Genty (Nevis)
REU Presentations
August 1, 2013
8 / 31
LArSoft - Cluster Studies - 1e − + 1p
0.3
PuritiesComb_fuzzy_cut
0.25
Entries
0.2
0.15
0.1
0.05
0
0
0.3
0.2
0.4
0.6
Purity
0.8
Entries
2747
0.2
0.15
0.1
0
0
1
EfficienciesComb_fuzzy_cut
0.3
EfficienciesComb_fuzzy_cut
Entries
0.2
0.2
0.15
0.1
0.05
0.4
0.6
Purity
0.8
1
EfficienciesComb_db_cut
EfficienciesComb_db_cut
0.25
2897
Frequency
Frequency
PuritiesComb_db_cut
0.05
0.25
0
0
PuritiesComb_db_cut
0.25
2897
Frequency
Frequency
0.3
PuritiesComb_fuzzy_cut
Entries
2747
0.2
0.15
0.1
0.05
0.2
0.4
0.6
Efficiency
Genty (Nevis)
0.8
1
REU Presentations
0
0
0.2
0.4
0.6
Efficiency
0.8
August 1, 2013
1
9 / 31
LArSoft - Energy Studies
Energy Fraction
Counts vs. Energy
5
Events
ADC Counts ×106
4
3
2
1
0
1
2
3
4
True e − Energy (GeV)
5
Visible energy fraction ∼ 45%
Important for detector calibration
1 GeV Electron
0.2
0.4
0.6
0.8
140
Entries
Mean
120
RMS
1
1000
0.411
0.03992
80
0.5 - 5.0 GeV Electron
60
40
20
0
0
Genty (Nevis)
Entries
1000
Mean
0.431
RMS 0.01129
100
Reconstructed ADC counts from
hits scaled linearly with true e −
energy
Events
0
450
400
350
300
250
200
150
100
50
0
0
REU Presentations
0.2
0.4
0.6
0.8
Ionization/True Energy
August 1, 2013
1
10 / 31
MicroBooNE Optical System
Phototube array
32-40, 8-inch photomultiplier array located behind TPC wireplanes will
collect Argon scintillation
The primary importance of the optical systems is for triggering on events
Optical information can also contribute to event reconstruction
I tested a R5912 8-inch PMT, similar to the ones used in MicroBooNE minus
the wavelength shifting coating and single coaxial input. Will be used to
study read out electronics
Genty (Nevis)
REU Presentations
August 1, 2013
11 / 31
PMT - Gain
Definition
Phototube gain is the ratio of secondary electrons collected on the anode to
primary electrons ejected from cathode → amplification factor
Procedure
1
Pulse PMT with blue LED @ 100 Hz
2
Record mean (µv ) peak height and
standard deviation
(σv ) of output
R
voltages, and Vdt over 6000
triggers
3
Repeat for different input voltages
G≡
Ns
Np
µv = CGNp
p
σv = CG Np
2
⇒ Np = (µv /σv )
and
R
Vdt
eR
R
2
Vdt σv
⇒G =
eR
µv
Ns =
G : Gain
Ns : Number of secondary electrons
Np : Number of primary electrons
Genty (Nevis)
REU Presentations
August 1, 2013
12 / 31
PMT - Gain - Results I
7
6
Gain vs. Voltage
100
(100 mV, 20 ns)/DIV
(200 mV, 20 ns)/DIV
(300 mV, 20 ns)/DIV
80
Primary Electrons vs. Voltage
(100 mV, 20 ns)/DIV
(200 mV, 20 ns)/DIV
(300 mV, 20 ns)/DIV
Gain ×107
5
60
Np
4
3
40
2
20
1
0
1100 1200 1300 1400 1500 1600 1700 1800
Voltage (V)
0
1100 1200 1300 1400 1500 1600 1700 1800
Voltage (V)
Took data at different oscilloscope
precisions (window size)
Number of primary electrons
deviates as function of input voltage
Spec. sheets reports gains at 107
Should remain constant
Optimal operating voltage is 1500 V
Photocathode electrons
non-poissonian?
Interesting gain response at high
voltages
Genty (Nevis)
REU Presentations
August 1, 2013
13 / 31
PMT - Gain - Results II
Gain vs. Time
7
7
5
Gain ×107
6
5
Gain ×107
6
Gain vs. Voltage
Measurement
Average
2
2
1
1
4
4
3
3
0
0
5
10
15
20 25 30
Time (min)
35
40
45
0
1100 1200 1300 1400 1500 1600 1700 1800
Voltage (V)
Variation in gain at constant
1500 V over 40 minutes
Spread is about ± one unit around
4 × 107
Genty (Nevis)
REU Presentations
Every measurement over 1.5 week
period plotted in red, blue square is
the average as estimate of
systematic uncertainty
Largest source of systematic
uncertainty is the oscilloscope
precision
August 1, 2013
14 / 31
PMT Splitter - Ringing - Setup
A current test of MicroBooNE’s optical system is called Bo. Bo is a liquid argon
test chamber for MicroBooNE photomultipliers, cold electronics, high voltage
system and much more.
An issue arose during electronics testing with the splitter used to split the HV
input from the PMT signal, signal reflection observed in shaper
Vin
Vout
L
R
C1
C2
A simple circuit was used to study the PMT signal reflection between the
splitting capacitor C2 and the PMT base
Genty (Nevis)
REU Presentations
August 1, 2013
15 / 31
PMT Splitter - Ringing - Reflection
Why is there reflection?
Impedance differentials along the length of the circuit reflect EM signals
Splitting circuit, and 50 Ω cable are at different impendances.
Toy Circuit
Varying L controls the timescale of reflection
Varying C2 controls amplitude
No ringing is observed when:
τcircuit = Rcable C2 τtravel =
L
vsignal
**Much greater ∼ 3-5 times
vsignal = 1 foot/1.5 ns
L = 4 → 20 meters
C2 = 1 nF → 10 nF
Genty (Nevis)
REU Presentations
August 1, 2013
16 / 31
PMT Splitter - Ringing - Tests
Short Cable - Shaper
30
25
25
20
20
Voltage (mV)
Voltage (mV)
30
15
10
5
Long Cable - Shaper
15
10
5
0
0
-5
-0.5
0
0.5
Time (µs)
1
1.5
-5
-0.5
0
0.5
Time (µs)
1
1.5
τcircuit = 50 Ω · 1 nF = 50 ns
Short cable L = 4 m
τcircuit > τtravel = 4 m · 1.5 ns/foot ∼ 20 ns → no ringing
Long cable L = 20 m
τcircuit ≯ τtravel = 10 m · 1.5 ns/foot ∼ 100 ns → yes ringing
Genty (Nevis)
REU Presentations
August 1, 2013
17 / 31
PMT Splitter - Ringing - Tests II
Increase τcircuit by C2 → 10 nF
Short Cable - Scope
Short Cable - Shaper
40
Voltage (mV)
Voltage (mV)
40
30
20
10
0
20
10
0
-0.5
0
0.5
1
Time (×100ns)
1.5
-0.5
Long Cable - Scope
30
0
0.5
Time (µs)
1
1.5
Long Cable - Shaper
40
Voltage (mV)
25
Voltage (mV)
30
20
15
10
5
0
30
20
10
0
-5
-0.5
0
0.5
1
1.5
Time (×100ns)
Genty (Nevis)
2
2.5
REU Presentations
-0.5
0
0.5
Time (µs)
1
August 1, 2013
1.5
18 / 31
PMT Splitter - Ringing - Results
Results
Bo circuit sees ringing in the shaper output when it shouldn’t, with same
parameters are test circuit
Bo circuit has another capacitor in series with the splitting capacitance
reducing effective capacitance
Bo circuit has high voltage across the splitting capacitance further reducing
capacitance
room temperature
||
10nF
10M
10k
PMT
500
GND
10k
10nF
(2kV)
10nF
(2kV)
450
16k
Anode
x1
Anode
x0.1
+HV in
GND
Capacitance in MicroBooNE splitter circuit used with Bo is being increased!
Genty (Nevis)
REU Presentations
August 1, 2013
19 / 31
PMT - Data Acquisition - Setup
Procedure
Use the controller module to trigger
a pulse generator
Feed the pulse to the RC circuit
built for the ringing tests. This
generates a narrow (few
nanosecond) PMT-like pulse of
variable charge depending on the
pulse amplitude.
Feed into the shaper and read out
through the FEM
Trigger Module
PC
Pulse
Generator
Decoder & Analysis Module
Beam Gate
Shaper
pmtbaseline
- written with Kazu’s framework
RC
Circuit
FEM
Ch. X
Pedestal calculation and subtraction
per shaper channel
Calculate signal peak and area for
pulse recon.
Genty (Nevis)
REU Presentations
August 1, 2013
20 / 31
PMT - Data Acquisition - Results
Pedestal mean and standard deviation calculated from the first 5 points of the
beam gate sample. Mean, RMS plotted versus FEM channel number
Pedestal Mean
RMS x
RMS y
20
7000
1.4
6000
1.2
5000
1
5000
0.8
4000
3000
0.6
3000
2000
0.4
2000
1000
0.2
1000
3.248
2050
4000
2045
2040
0
1.6
2049
11.54
5
10
15
20 25
Channel
30
35
40
Entries 400000
Mean x
20
Mean y 0.3674
RMS x
11.54
RMS y 0.1782
8000
ADC
Mean x
Mean y
2055
ADC
Pedestal RMS
Entries 400000
2060
0
0
0
8000
7000
6000
5
10
15
20 25
Channel
30
35
40
0
Pedestal mean ∼ 2049, pedestal varies over 10 ADC counts
Pedestal RMS ∼ 0.37
Genty (Nevis)
REU Presentations
August 1, 2013
21 / 31
PMT - Data Acquisition - Results II
Integrated Charge Ch. 4
1000
Distribution of integrated charge.
Increasing in amplitude to the right
by 500 mV input. High energy tail
on each distribution → not gaussian
Pulse Count
800
600
400
200
100 150 200 250 300 350 400 450 500 550 600
Charge
Peak Heights Ch. 4
2200
2000
2/9 distribution of peak heights,
another estimator of pulse energy.
Would expect peak heights to be
gaussian as well but because of a
digitization effect there are 3
distributions
1800
1600
Events
1400
1200
1000
800
600
400
200
0
70
72
74
76
78
Peak Amplitude
Genty (Nevis)
80
82
REU Presentations
August 1, 2013
22 / 31
PMT - Data Acquisition - Results III
Charge: Amp=4.0
Pulse Height Division
900
3500
800
700
Pulse Count
Pulse Count
3000
2500
2000
1500
600
500
400
300
1000
200
500
100
0
24
25
26
27
28
Pulse Amplitude
29
0
140
30
Make 3 cuts on pulse amplitude
distribution
145
150
155
160 165
Charge
170
175
180
185
Charge distribution split by peak
cuts
Find max bin, look ±0.5 as estimate
of digitization effect
Genty (Nevis)
REU Presentations
August 1, 2013
23 / 31
PMT - Data Acquisition - Results IV
Middle Charge Sum Ch. 4
450
/ ndf
p0
p1
255.1 / 7
-12.28 ± 0.2447
43.64 ± 0.02926
5
400
All
4
350
χ2
Middle Mean Fitted Charge
500
χ2 vs. Amplitude Ch. 4
6
χ2
3
300
2
250
200
Middle
1
150
4
5
6
7
8
9
Amplitude
10
11
12
0
4
5
6
7
8
9
Amplitude
10
11
12
Magenta: middle peak fits
Blue: fits without cuts
For each input A, plot mean
For each input A, plot χ2 goodness
of fit parameter
Linear as function of input A
Cuts are indication better selection
of sample selection
Genty (Nevis)
REU Presentations
August 1, 2013
24 / 31
PMT - Data Acquisition - Results V
Repeat over all shaper channels
48
-5
-10
46
-15
44
-20
42
40
Y-intercept Fitted vs. Channel
0
50
Fitted Y-intercept Charge
Slope of Fitted Mean Charge
Mean Fitted Charge vs. Channel
-25
0
2
4
6
8
10 12
Shaper Channel Number
-30
14
Slope of fitted mean plotted over
channel number
0
2
4
6
8
10 12
Shaper Channel Number
14
Y-intercept of fitted mean plotted
over channel number
Shows non linearity at low energy
(A< 4)
Genty (Nevis)
REU Presentations
August 1, 2013
25 / 31
Thanks To:
David Caratelli for being a great lab partner
Kazu for teaching me PyROOT
Mike Shaveitz for the opportunity to work on
MicroBooNE
John Parsons for adminstrating the REU
Georgia for getting me started with LArSoft
REU Students for being great friends
Genty (Nevis)
REU Presentations
August 1, 2013
26 / 31
Backup Slides
BACKUP
Genty (Nevis)
REU Presentations
August 1, 2013
27 / 31
Fermilab Booster
A
bs
or
be
r
Producing Neutrinos - Fermilab Booster
Decay Pipe
Dirt
Detector
50 m
Target and
Horn
1
2
3
4
8 GeV protons produced in booster
Impinge on Beryllium target, magnetic horn focusses π ± & K ± depending on
neutrino mode
Mesons decay via → µ± + ν̄µ /νµ channel, some µ± → e ± + ν̄µ /νµ + νe /ν̄e
Absorber filters charged leptons
Genty (Nevis)
REU Presentations
August 1, 2013
28 / 31
MicroBooNE - Advantages
Liquid argon TPCs have a low
energy resolution at a few MeV, far
below the hundreds of MeV
threshold on MiniBooNE, and will
be able to resolve the size of the
signal at lower energies.
MiniBooNE could not differentiate
between electrons and photons, a
TPC can “see” the difference → e −
connected to a primary vertex which
is singly ionizing, γ are doubly
ionizing and have a gap between
vertex
Detector R&D for larger TPC
experiments to search for CP
violation in neutrino sector
Genty (Nevis)
TPC wireplanes: red and green “induction”
planes ±60◦ to vertical, Blue parallel
“collection” plane
REU Presentations
August 1, 2013
29 / 31
LArSoft - Cluster Studies
1
2
3
4
Generate single electron, muon and uniform flux CC νe events with singles.fcl
and GENIE. Filter for 1e − + 1p final states
Reconstruct clusters with modified uboone offline .fcl script
Feed to a module I wrote, MCHitter, to calculate purity and efficiency of
reconstructed clusters
Compare DBSCAN, FuzzyCluster
⇒
Genty (Nevis)
REU Presentations
August 1, 2013
30 / 31
LArSoft - Cluster Studies - Purity & Efficiency
Efficiency
=
# of hits from trackID in cluster
total # of hits for that trackID
Purity
=
# of hits from trackID in cluster
total # of hits in cluster
e−
e−
π−
π−
Measures
Measures
How much of a cluster is composed
of a each true particle
How many of all hits the particle
generated are in a specific cluster
If less than 1: clustering algorithm
could not distinguish true particle
hits from one another
If less than 1: algorithm failed to
group the hits created by the
particle into a single cluster
Genty (Nevis)
REU Presentations
August 1, 2013
31 / 31
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement