T Pulse Identification Method for Overlapped Pulses · Figure 16 shows an event display for two...
1
1 1.5 2 2.5 3 3.5 4 -5 10 -4 10 -3 10 -2 10 -1 10 1 0 10 20 30 40 50 60 70 80 90 100 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Time [ns] 0 100 200 300 400 500 Pulse Height [ADC counts] 600 650 700 750 800 850 900 950 1000 0 10 20 30 40 50 60 70 80 90 100 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 40ns 20mV PMT signal (CsI) recorded by an oscilloscope Recorded waveform at ADC module after filter 11 division of pulse heights 1 10 2 10 3 10 2 10 3 10 4 10 0 2 4 6 8 10 12 14 16 18 20 hchi2ndf_height_det_4 Entries 1679733 Mean x 643.8 Mean y 0.8222 RMS x 650.6 RMS y 0.5894 PulseHeight[ ADCcounts] χ2/NDF Pulse Identification Method for Overlapped Pulses O K T ν ν s d Search for • Rere Decay: BR=2.4x10 -11 (theoretical prediction) • Direct CP violation process. • Br(K 0 L →π 0 νν)∝η2 • Small theoretical uncertainty : 1~2% → Golden Mode to Test Standard Model • Sensitive to new Physics through loop diagrams → Good Probe for New Physics K 0 L → π 0 ν ν 576 pure CsI c K L K π 0 cay r K L K π 0 cay r Feynman diagrams Yasuyuki Sugiyama, Kazufumi Sato, Manabu Togawa, Taku Yamanaka - Osaka University Eito Iwai - High Energy Accelerator Research Organization FB NCC MB CV CsI calorimeter CC03 OEV CC04 CC05 CC06 BHCV BHPV LCV BCV HINEMOS K L π 0 →2γ ν ν Detect 2γ (π 0 →2γ) and nothing else • detect 2γ by CsI Calorimeter • reconstruct the π 0 decay vertex from energy and timing. • Confirm nothing else from the decay. • Cover the decay region with high efficiency hermetic veto counters. First Physics Run in May 2013 @J-PARC Requirement for pulse identification method J-PARC KOTO experiment Physics Motivation Experimental setup and method Conclusion We established a method to discriminate pulses from waveform using pulse shape template. Overlapped pulses with distance more than 30 ns can be separated. Pulse Identification with Template Fitting Event with Overlapped Pulses in May 2013 run(MB) • Low inefficiency for particle detection is required for veto. • Multiple pulses in the veto timing window can be a source of inefficiency. • Recording waveform with 14bit 125MHz ADC. • shape the signals into broad gaussian shapes with a 10-pole Bessel Filter. • Analyze the waveform and identify pulses. • Need to establish the method to identify pulses -10 0 10 20 30 40 0 0.2 0.4 0.6 0.8 1 detID 80 Region 7 detID 80 Region 7 0 10 20 30 40 50 60 500 0 10 20 30 40 50 60 500 600 700 800 900 1000 1100 1200 0 10 20 30 40 50 60 500 600 700 800 900 1000 1100 1200 Average after Alignment, Normalization Time[clock = 8ns] Pulse Height [arbitrary unit] Template waveform Pulse Height [ADC counts] • Use “template”: typical pulse shape for fitting. • Prepare template for each detector. • separate template for 11 regions of pulse height. • Make template by averaging the waveform from many events. • Start from fitting with 1pulse assumption. • If χ2/NDF>threshold→ Assume one more pulse to fit and try fitting. • Use peak of (Data- Fit)/Error as candidate of additional peak. • Iterate this procedure until χ2/NDF becomes low enough. Pulse Identification with template fitting Template for fitting Performance of pulse identification 0 10 20 30 40 50 60 3 10 / ndf 2 r 41.03 / 60 p0 0 ± 500 p1 0 ± 2 p2 0 ± 0 p3 0 ± 0 p4 1.968 ± 5175 p5 0.00354 ± 13.48 p6 6.889 ± 2213 p7 0.01602 ± 21.89 0 10 20 30 40 50 60 3 10 / ndf 2 r 1.118e+05 / 62 p0 0 ± 500 p1 0 ± 1 p2 0 ± 0 p3 0 ± 0 p4 1.39 ± 5250 p5 0.002687 ± 13.75 0 10 20 30 40 50 60 -40 -20 0 20 40 60 80 100 120 :True Peak position (Data-Fit)/Error χ2/NDF=1800 Time[clock = 8ns] Pulse Height [ ADC counts] Time[clock = 8ns] Pulse Height [ ADC counts] Pulse Height [ ADC counts] Time[clock = 8ns] χ2/NDF=0.68 Example with simulated two pulses Fitting with 1 pulse assumption Fitting with 2 pulse assumption Time [Clock=8ns] 0 10 20 30 40 50 60 Pulse Height [ADC counts] 300 350 400 450 500 Main Pulse From Physics Triggered event Time [Clock=8ns] 0 10 20 30 40 50 60 Pulse Height [ADC counts] 400 600 800 1000 1200 Pulse From Random Trigger Data Time [Clock=8ns] 0 10 20 30 40 50 60 Pulse Height [ADC counts] 400 600 800 1000 1200 CFTime: before:27.089 after7.36957 Time [Clock=8ns] 0 10 20 30 40 50 60 Pulse Height [ADC counts] 400 600 800 1000 1200 / ndf 2 r 24.63 / 59 p0 1.042 ± 311.3 p1 0 ± 2 p2 0 ± 0 p3 0 ± 60 p4 1.903 ± 917 p5 0.01271 ± 11.11 p6 2.731 ± 205.4 p7 0.06433 ± 30.93 fittime before:30.9141,after30.9278 Main pulse from Physics trigger Merge Pulse from random trigger Fit Evaluate performance with merging two waveforms with single pulse • Generate waveform of overlapped pulses from data taken in the physics run. • Merge two waveforms of event with single pulse from Physics and Random trigger data. • Select single pulse event by requiring χ2/NDF<2 for fitting with single pulse assumption. • Set threshold for χ2/NDF at 5 for Pulse Identification. • Evaluate performance for the waveform of “Main Barrel (MB)” χ2/NDF for fitting cosmic-ray event with 1pulse assumption (MB) x: distance between 2 pulses[ns] y: Log10 (Height ratio: Late pulse /Early pulse ) x: distance between 2pulses[ns] y: Separation Efficiency for MB pulses 2 pulse separation efficiency against pulse height and timing (for the case with pulse height of both pulse >100 ADC counts~2MeV) x: Log10(Main pulse height) y: Inefficiency Blue: 1 Pulse assumption Red: Pulse Identification Evaluate Inefficiency due to accidental pulse overlapping • Compare timing of main pulse before and after merging random trigger pulses. • Inefficient events: events with timing shifted more than 30 ns after merging. • Compare inefficiency between timing with 1pulse assumption or pulse identification. z: Separation Efficiency for MB pulses MB
Transcript of T Pulse Identification Method for Overlapped Pulses · Figure 16 shows an event display for two...
h_badevt_height_det_4Entries 2.444672e+07Mean 1.29RMS 0.2173
1 1.5 2 2.5 3 3.5 4
-510
-410
-310
-210
-110
1
h_badevt_height_det_4Entries 2.444672e+07Mean 1.29RMS 0.2173
nbadevt vs height for MB (window = +-30ns)
h_goodevt_time_det_4Entries 766981Mean 62.67RMS 22.96
0 10 20 30 40 50 60 70 80 90 1000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
h_goodevt_time_det_4Entries 766981Mean 62.67RMS 22.96ngoodevt time for MB
Time [ns]0 100 200 300 400 500
Puls
e H
eigh
t [AD
C c
ount
s]
600
650
700
750
800
850
900
950
1000Graph
0 10 20 30 40 50 60 70 80 90 100-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
h_goodevt_height_det_4Entries 15738Mean x 63.5Mean y 0.1873RMS x 21.7RMS y 0.8464
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
h_goodevt_height_det_4Entries 15738Mean x 63.5Mean y 0.1873RMS x 21.7RMS y 0.8464
ngoodevt time vs height for MB
40ns
20mV
PMT signal (CsI) recorded by an oscilloscope
Recorded waveform at ADC module after filter
11 division of pulse heights
1
10
210
310
210 310 4100
2
4
6
8
10
12
14
16
18
20
chi2/ndf vs height MBhchi2ndf_height_det_4
Entries 1679733Mean x 643.8Mean y 0.8222RMS x 650.6RMS y 0.5894
chi2/ndf vs height MB
PulseHeight[ ADCcounts]
χ2/NDF
Pulse Identification Method for Overlapped PulsesOK Tν
νs
d
Search for !
• Rere Decay: BR=2.4x10-11
(theoretical prediction) • Direct CP violation process. • Br(K0L→π0νν)∝η2 • Small theoretical uncertainty : 1~2% → Golden Mode to Test Standard Model !
• Sensitive to new Physics through loop diagrams → Good Probe for New Physics
Another background is π0 production from beam neutrons interactingwith residual gas in the decay region. In order to suppress this background,the decay volume is evacuated to 10−5 Pa, as was obtained in E391a byseparating the detector and the decay region with a thin film.
Figure 14: Schematic view of detector setup.
4.3.1 Calorimeter
The electromagnetic calorimeter measures the positions and energies of pho-tons to reconstruct π0 in the K0
L → π0νν decay. In the E391a experiment,the Calorimeter was made of 576 pure CsI crystals. Each crystal was 7.0×7.0cm2 and 30-cm long (16 X0) [59].
For the experiment at J-PARC, we plan to replace these crystals with thepure CsI crystals used in the calorimeter of the Fermilab KTeV experiment.The crystals, called “KTeV CsI crystals” hereafter, are smaller in the crosssection and longer in the beam direction (50 cm, 27 X0) than the crystals inE391a, which ensures us much better performance in the new experiment.Figure 15 shows the layout of the new Calorimeter, which then consists of2576 crystals. These crystals are of two sizes, 2.5 × 2.5 × 50 cm3 for thecentral region (2240 blocks), and 5.0×5.0×50 cm3 for the outer region (336blocks) of the Calorimeter.
The reasons for replacing the calorimeter are as follows.
• Reduce the probability of missing photons due to fused clusters.If two photons hit the Calorimeter close to each other, the generatedshowers will overlap and be misidentified as a single photon. Figure 16shows an event display for two photons that enter the CsI Calorimeterwith 6-cm separation. By using the KTeV CsI crystals, two photons
28
4.7.3 KL → π0π0π0 background . . . . . . . . . . . . . . . . 604.7.4 KL → π+π−π0 Background . . . . . . . . . . . . . . . 624.7.5 KL → π−e+ν Background . . . . . . . . . . . . . . . . 634.7.6 Neutron Background . . . . . . . . . . . . . . . . . . . 654.7.7 Accidental halo neutron background . . . . . . . . . . 67
5 Step 2 695.1 New beamline . . . . . . . . . . . . . . . . . . . . . . . . . . . 695.2 Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715.3 Sensitivity and Background . . . . . . . . . . . . . . . . . . . 725.4 Detector R&D . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.4.1 Endcap Calorimeter . . . . . . . . . . . . . . . . . . . 745.4.2 Charged Veto . . . . . . . . . . . . . . . . . . . . . . . 755.4.3 Pipeline readout . . . . . . . . . . . . . . . . . . . . . 78
6 Schedule and Cost 796.1 Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 796.2 Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7 Conclusion 81
A Appendices 82A.1 Beam Simulation . . . . . . . . . . . . . . . . . . . . . . . . . 82A.2 Photon Veto Inefficiency . . . . . . . . . . . . . . . . . . . . . 87A.3 Charged Particle Veto Inefficiency . . . . . . . . . . . . . . . 90A.4 Expected Performance of Beam Hole Photon Veto . . . . . . 93
4
Another background is π0 production from beam neutrons interactingwith residual gas in the decay region. In order to suppress this background,the decay volume is evacuated to 10−5 Pa, as was obtained in E391a byseparating the detector and the decay region with a thin film.
Figure 14: Schematic view of detector setup.
4.3.1 Calorimeter
The electromagnetic calorimeter measures the positions and energies of pho-tons to reconstruct π0 in the K0
L → π0νν decay. In the E391a experiment,the Calorimeter was made of 576 pure CsI crystals. Each crystal was 7.0×7.0cm2 and 30-cm long (16 X0) [59].
For the experiment at J-PARC, we plan to replace these crystals with thepure CsI crystals used in the calorimeter of the Fermilab KTeV experiment.The crystals, called “KTeV CsI crystals” hereafter, are smaller in the crosssection and longer in the beam direction (50 cm, 27 X0) than the crystals inE391a, which ensures us much better performance in the new experiment.Figure 15 shows the layout of the new Calorimeter, which then consists of2576 crystals. These crystals are of two sizes, 2.5 × 2.5 × 50 cm3 for thecentral region (2240 blocks), and 5.0×5.0×50 cm3 for the outer region (336blocks) of the Calorimeter.
The reasons for replacing the calorimeter are as follows.
• Reduce the probability of missing photons due to fused clusters.If two photons hit the Calorimeter close to each other, the generatedshowers will overlap and be misidentified as a single photon. Figure 16shows an event display for two photons that enter the CsI Calorimeterwith 6-cm separation. By using the KTeV CsI crystals, two photons
28
4.7.3 KL → π0π0π0 background . . . . . . . . . . . . . . . . 604.7.4 KL → π+π−π0 Background . . . . . . . . . . . . . . . 624.7.5 KL → π−e+ν Background . . . . . . . . . . . . . . . . 634.7.6 Neutron Background . . . . . . . . . . . . . . . . . . . 654.7.7 Accidental halo neutron background . . . . . . . . . . 67
5 Step 2 695.1 New beamline . . . . . . . . . . . . . . . . . . . . . . . . . . . 695.2 Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715.3 Sensitivity and Background . . . . . . . . . . . . . . . . . . . 725.4 Detector R&D . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.4.1 Endcap Calorimeter . . . . . . . . . . . . . . . . . . . 745.4.2 Charged Veto . . . . . . . . . . . . . . . . . . . . . . . 755.4.3 Pipeline readout . . . . . . . . . . . . . . . . . . . . . 78
6 Schedule and Cost 796.1 Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 796.2 Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7 Conclusion 81
A Appendices 82A.1 Beam Simulation . . . . . . . . . . . . . . . . . . . . . . . . . 82A.2 Photon Veto Inefficiency . . . . . . . . . . . . . . . . . . . . . 87A.3 Charged Particle Veto Inefficiency . . . . . . . . . . . . . . . 90A.4 Expected Performance of Beam Hole Photon Veto . . . . . . 93
4
Another background is π0 production from beam neutrons interactingwith residual gas in the decay region. In order to suppress this background,the decay volume is evacuated to 10−5 Pa, as was obtained in E391a byseparating the detector and the decay region with a thin film.
Figure 14: Schematic view of detector setup.
4.3.1 Calorimeter
The electromagnetic calorimeter measures the positions and energies of pho-tons to reconstruct π0 in the K0
L → π0νν decay. In the E391a experiment,the Calorimeter was made of 576 pure CsI crystals. Each crystal was 7.0×7.0cm2 and 30-cm long (16 X0) [59].
For the experiment at J-PARC, we plan to replace these crystals with thepure CsI crystals used in the calorimeter of the Fermilab KTeV experiment.The crystals, called “KTeV CsI crystals” hereafter, are smaller in the crosssection and longer in the beam direction (50 cm, 27 X0) than the crystals inE391a, which ensures us much better performance in the new experiment.Figure 15 shows the layout of the new Calorimeter, which then consists of2576 crystals. These crystals are of two sizes, 2.5 × 2.5 × 50 cm3 for thecentral region (2240 blocks), and 5.0×5.0×50 cm3 for the outer region (336blocks) of the Calorimeter.
The reasons for replacing the calorimeter are as follows.
• Reduce the probability of missing photons due to fused clusters.If two photons hit the Calorimeter close to each other, the generatedshowers will overlap and be misidentified as a single photon. Figure 16shows an event display for two photons that enter the CsI Calorimeterwith 6-cm separation. By using the KTeV CsI crystals, two photons
28
Feynman diagrams
Yasuyuki Sugiyama, Kazufumi Sato, Manabu Togawa, Taku Yamanaka - Osaka University Eito Iwai - High Energy Accelerator Research Organization
FB NCC MB CVCsI calorimeter
CC03OEV
CC04 CC05 CC06 BHCV BHPV
LCVBCVHINEMOS
Saturday, April 20, 2013
KL
π0→2γ
ν ν
Detect 2γ (π0→2γ) and nothing else • detect 2γ by CsI Calorimeter
• reconstruct the π0 decay vertex from energy and timing. • Confirm nothing else from the decay.
• Cover the decay region with high efficiency hermetic veto counters.
First Physics Run in May 2013 @J-PARC
Requirement for pulse identification method
J-PARC KOTO experimentPhysics Motivation
Experimental setup and method
ConclusionWe established a method to discriminate pulses from waveform using pulse shape template.Overlapped pulses with distance more than 30 ns can be separated.
Pulse Identification with Template Fitting
Event with Overlapped Pulses in May 2013 run(MB)
• Low inefficiency for particle detection is required for veto. • Multiple pulses in the veto timing window can be a source of inefficiency. • Recording waveform with 14bit 125MHz ADC. • shape the signals into broad gaussian shapes
with a 10-pole Bessel Filter. • Analyze the waveform and identify pulses.
!
• Need to establish the method to identify pulses
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 0detID 80 Region 0
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 1detID 80 Region 1
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 2detID 80 Region 2
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 3detID 80 Region 3
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 4detID 80 Region 4
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 5detID 80 Region 5
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 6detID 80 Region 6
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 7detID 80 Region 7
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 8detID 80 Region 8
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 9detID 80 Region 9
-10 0 10 20 30 40
0
0.2
0.4
0.6
0.8
1
detID 80 Region 10detID 80 Region 10
0 10 20 30 40 50 60
500
600
700
800
900
1000
1100
1200
Graph
0 10 20 30 40 50 60
500
600
700
800
900
1000
1100
1200
Graph
0 10 20 30 40 50 60
500
600
700
800
900
1000
1100
1200
Graph
Average after
Alignment,Normalization
Time[clock = 8ns]
Pulse Height [arbitrary unit]
Template waveform
Pulse Height [ADC
counts]
• Use “template”: typical pulse shape for fitting. • Prepare template for each detector. • separate template for 11 regions of pulse height. • Make template by averaging the waveform from many events.
• Start from fitting with 1pulse assumption. • If χ2/NDF>threshold→ Assume one more pulse to fit and try fitting. • Use peak of (Data- Fit)/Error as candidate of additional peak.
• Iterate this procedure until χ2/NDF becomes low enough.
Pulse Identification with template fitting
Template for fitting
Performance of pulse identification
0 10 20 30 40 50 60
310
/ ndf 2r 1.118e+05 / 62p0 0± 500 p1 0± 1 p2 0± 0 p3 0± 0 p4 1.39± 5250 p5 0.002687± 13.75
/ ndf 2r 1.118e+05 / 62p0 0± 500 p1 0± 1 p2 0± 0 p3 0± 0 p4 1.39± 5250 p5 0.002687± 13.75
truetime=13.47,21.8099 trueene=797.046(height=5140.94) ,344.785(height=2223.86)
0 10 20 30 40 50 60
310
/ ndf 2r 41.03 / 60p0 0± 500 p1 0± 2 p2 0± 0 p3 0± 0 p4 1.968± 5175 p5 0.00354± 13.48 p6 6.889± 2213 p7 0.01602± 21.89
/ ndf 2r 41.03 / 60p0 0± 500 p1 0± 2 p2 0± 0 p3 0± 0 p4 1.968± 5175 p5 0.00354± 13.48 p6 6.889± 2213 p7 0.01602± 21.89
npeak=2
0 10 20 30 40 50 60-40
-20
0
20
40
60
80
100
120
Data - (1st trial fitting function) peak candidate:23
0 10 20 30 40 50 60
310
/ ndf 2r 1.118e+05 / 62p0 0± 500 p1 0± 1 p2 0± 0 p3 0± 0 p4 1.39± 5250 p5 0.002687± 13.75
/ ndf 2r 1.118e+05 / 62p0 0± 500 p1 0± 1 p2 0± 0 p3 0± 0 p4 1.39± 5250 p5 0.002687± 13.75
truetime=13.47,21.8099 trueene=797.046(height=5140.94) ,344.785(height=2223.86)
0 10 20 30 40 50 60
310
/ ndf 2r 41.03 / 60p0 0± 500 p1 0± 2 p2 0± 0 p3 0± 0 p4 1.968± 5175 p5 0.00354± 13.48 p6 6.889± 2213 p7 0.01602± 21.89
/ ndf 2r 41.03 / 60p0 0± 500 p1 0± 2 p2 0± 0 p3 0± 0 p4 1.968± 5175 p5 0.00354± 13.48 p6 6.889± 2213 p7 0.01602± 21.89
npeak=2
0 10 20 30 40 50 60-40
-20
0
20
40
60
80
100
120
Data - (1st trial fitting function) peak candidate:23
:True Peak position
(Data-Fit)/Error
χ2/NDF=1800
Time[clock = 8ns]
Pulse Height[ ADC
counts]
Time[clock = 8ns]
Pulse Height[ ADC
counts]
Pulse Height[ ADC
counts]
Time[clock = 8ns]
χ2/NDF=0.68
Example with simulated two pulses
Fitting with 1 pulse
assumption
Fitting with 2 pulse assumption
Time [Clock=8ns]0 10 20 30 40 50 60
Puls
e H
eigh
t [AD
C c
ount
s]
300
350
400
450
500
Main Pulse From Physics Triggered event
Time [Clock=8ns]0 10 20 30 40 50 60
Puls
e H
eigh
t [AD
C c
ount
s]
400
600
800
1000
1200
Pulse From Random Trigger Data
Time [Clock=8ns]0 10 20 30 40 50 60
Puls
e H
eigh
t [AD
C c
ount
s]
400
600
800
1000
1200
CFTime: before:27.089 after7.36957
Time [Clock=8ns]0 10 20 30 40 50 60
Puls
e H
eigh
t [AD
C c
ount
s]
400
600
800
1000
1200
/ ndf 2r 24.63 / 59p0 1.042± 311.3 p1 0± 2 p2 0± 0 p3 0± 60 p4 1.903± 917 p5 0.01271± 11.11 p6 2.731± 205.4 p7 0.06433± 30.93
/ ndf 2r 24.63 / 59p0 1.042± 311.3 p1 0± 2 p2 0± 0 p3 0± 60 p4 1.903± 917 p5 0.01271± 11.11 p6 2.731± 205.4 p7 0.06433± 30.93
fittime before:30.9141,after30.9278
Main pulsefrom
Physicstrigger
Merge
Pulse from
randomtrigger
Fit
Evaluate performance with merging two waveforms with single pulse • Generate waveform of overlapped pulses from data taken in the physics run. • Merge two waveforms of event with single pulse from Physics and Random
trigger data. • Select single pulse event by requiring χ2/NDF<2 for fitting with
single pulse assumption. • Set threshold for χ2/NDF at 5 for Pulse Identification. • Evaluate performance for the waveform of “Main Barrel (MB)”
χ2/NDF for fitting cosmic-ray event with 1pulse assumption
(MB)x: distance between 2 pulses[ns]
y: Log10 (Height ratio: Late pulse
/Early pulse )
x: distance between 2pulses[ns]
y: Separation Efficiency for MB pulses
2 pulse separation efficiency against pulse height and timing (for the case with pulse height of both pulse >100 ADC counts~2MeV)
x: Log10(Main pulse height)
y: Inefficiency
Blue: 1 Pulse assumption Red: Pulse Identification
Evaluate Inefficiency due to accidental pulse overlapping • Compare timing of main pulse before and
after merging random trigger pulses. • Inefficient events: events with timing
shifted more than 30 ns after merging. • Compare inefficiency between timing with
1pulse assumption or pulse identification.
z: Separation Efficiency for MB pulses
MB
![Toward Robust Recommendation Systems for Scholarly Papers and Mobile … › ~sugiyama › papers › Talk-Indian… · [Sugiyama and Kan, JCDL’10] References Citation (cit) papers](https://static.fdocuments.us/doc/165x107/5f1f299fd1909c50552b2e90/toward-robust-recommendation-systems-for-scholarly-papers-and-mobile-a-sugiyama.jpg)
