Hit Primitives & Basic Clustering for Supernova Triggering - - PowerPoint PPT Presentation

Hit Primitives & Basic Clustering for Supernova Triggering

Alexander Booth

1

DAQ Sim Weekly Meeting. December 18, 2017

• Triggering based on individual hits.
• Triggering based on clusters in channel and time.

Overview:

Analysis Details

2

GOAL: Understand supernova triggering efficiencies and corresponding background rates for different levels of trigger. Explore the potential to trigger on just the HIT PRIMITIVES of individual hits and compare this performance to CLUSTERING hits in channel and time.

• Using an amended version of the DAQSimAna (M. Baird, K. Warburton)

module in dunetpc.

• Running on files produced for the DUNE physics week, SN+radiologicals

+noise.

• Non-compressed, 1000 events each of 1 drift window and containing 1

MARLEY neutrino per event. Include Ar42. 1x6x2 geometry.

/pnfs/dune/persistent/users/talion/v06_56_00/reco/snb_bkg_nocompression_dune10kt_1x2x6/ files.list

• Gauss hit finder to pick out hits. All collection plane. Save hit primitives

such as hit time, ADC sum of hit, hit RMS etc. Why not fast hit? (For now).

• Backtrack each of these hits to a generator - was it radiological/noise/

supernova.

Why not fast hit?

3

Problems with Hit Size = End time - start time

• Fast hit looks for time bins with ADC above a user defined ADC threshold. >2

bins above threshold, calls it a hit.

• Start time = first bin above threshold, end time = last bin above threshold.

Results in many ‘skinny’ hits in time -> Many hits not correctly backtracked.

htemp

Entries 307076 Mean 1.409 Std Dev 2.083 1 2 3 4 5 6 7 8 9 10 Generator 20 40 60 80 100 120 140 160 180 200

3

10 × Number of Hits

htemp

Entries 307076 Mean 1.409 Std Dev 2.083

Generator Type, Fast Hit (20ADC)

htemp

Entries 396728 Mean 4.227 Std Dev 0.9519 1 2 3 4 5 6 7 8 9 10 Generator 50 100 150 200 250 300

3

10 × Number of Hits

htemp

Entries 396728 Mean 4.227 Std Dev 0.9519

Generator Type, Gauss Hit

Generator type 0 is ‘noise’.

Gauss Hit

4

Originally developed by Jonathan Asaadi but expanded by many.

• Finds pulses in each view above individually configured thresholds.
• ‘Touching’ hits on a channel are merged up to a configurable max.
• Hits fit to a gaussian peak for:
• Start and end time.
• Peak time.
• Peak ADC.
• Total hit ADC is integral of raw data, not fit by default.
• Default generous max Chi^2 for allowed hits.

Has a hard coded hit size minimum of 5 ticks.

Hit Primitives

5

1 2 3 4 5 6 7 8 HitRMS, (ticks) 1 10

2

10

3

10

4

10

5

10 Number of Hits

Comparing MARLEY, Radiologicals & Noise: HitRMS

MARLEY Radiologicals Noise

200 400 600 800 1000 HitSADC, (ADC) 1 10

2

10

3

10

4

10

5

10 Number of Hits

Comparing MARLEY, Radiologicals & Noise: HitSADC

MARLEY Radiologicals Noise

Marley v Radiologicals v Noise

6

20 40 60 80 100 HitPeak, (ADC) 1 10

2

10

3

10

4

10

5

10 Number of Hits

Comparing MARLEY, Radiologicals & Noise: HitPeak

MARLEY Radiologicals Noise

Some variables show fairly good separation. Potential to make pre-clustering cuts.

1 2 3 4 5 6 7 8 HitRMS, (ticks) 1 10

2

10

3

10

4

10

5

10 Number of Hits

Comparing MARLEY to Individual Radiologicals & Noise: HitRMS

Noise MARLEY APA frame, Co60 CPA fram, K40 Ar39 n Kr Po Rn Ar42

200 400 600 800 1000 HitSADC, (ADC) 1 10

2

10

3

10

4

10

5

10 Number of Hits

Comparing MARLEY to Individual Radiologicals & Noise: HitSADC

Noise MARLEY APA frame, Co60 CPA fram, K40 Ar39 n Kr Po Rn Ar42

Marley v Individual Generator

7

20 40 60 80 100 HitPeak, (ADC) 1 10

2

10

3

10

4

10

5

10 Number of Hits

Comparing MARLEY to Individual Radiologicals & Noise: HitPeak

Noise MARLEY APA frame, Co60 CPA fram, K40 Ar39 n Kr Po Rn Ar42

Ar42

1 2 3 4 5 6 7 8 9 Number of Supernova Like Hits 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Efficiency

Supernova Trigger Efficiency: HitPeak v HitRMS Cut

Marley: 69%

Supernova Trigger Efficiency: HitPeak v HitRMS Cut

1 2 3 4 5 6 7 8 9 Number of Supernova Like Hits 200 400 600 800 1000 Rate, (Hz)

Background Rate: HitSADC v HitRMS Cut

Radilogicals: 0.15% Noise: 0.34%

Background Rate: HitSADC v HitRMS Cut

Triggering on hit primitives

8

Cut:

• HitRMS<1.8TDC, HitPeak >=20ADC
• HitRMS>=1.8TDC, HitPeak>exp(-HitRMS+5)

% of hits left in the sample after cut applied. Define trigger: Number of SN like hits in an event. Background rate: Count number of triggers due to backgrounds. SN efficiency: Did we have a trigger due to Marley hits?

Clustering in Channels & Time

9

Clustering Algorithm

10

• Hits ordered sequentially by channel. Walk along the wires looking for hits on ‘adjacent’
• channels. Can modify adjacent channel tolerance (ACT).

Channel ACT

X X X X X X

1 2 3 4 5 6 7 8 9 10

• Calculate the total ADC sum of the hits in the cluster.
• Order hits sequentially in time within each cluster. Walk through hits, checking time
• separation. Can modify adjacent hit time separation (ATS).
• Look for a number of sequential time hits. Number of adjacent time hits.

If ACT = 2 X X X X Time X X X ATS X X X X X X

If required number of adjacent time hits > 2

PASS FAIL

• Can then cut on minimum number of channels in a cluster or cluster width.

1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 HitSADC Total, (/Cluster) 1 10

2

10

3

10

4

10

5

10 Number of Clusters

Total ADC Sum per Cluster

MARLEY Backgrounds

5 10 15 20 25 30 Number of Hits, (/Cluster) 1 10

2

10

3

10

4

10 Number of Clusters

Number of Hits per Cluster

MARLEY Backgrounds

5 10 15 20 25 30 Number of Channels 1 10

2

10

3

10

4

10 Number of Clusters

Number of Channels per Cluster

MARLEY Backgrounds

Cuts/Parameters

11

Anything in blue is a parameter that can be changed/cut on.

Efficiency and Background Rates

12

5 10 15 20 25 30 35 40 Minimum Channels With Hits in a Cluster (#channels) 0.2 0.4 0.6 0.8 1 Efficiency

Supernova Trigger Efficiency, Varying Minimum Number of Channels With Hits in a Cluster Supernova Trigger Efficiency, Varying Minimum Number of Channels With Hits in a Cluster

5 10 15 20 25 30 35 40 Minimum Channels With Hits in a Cluster (#channels) 10000 20000 30000 40000 50000 60000 70000 Rate, (Hz) Background Rate, Varying Minimum Number of Channels With Hits in a Cluster Background Rate, Varying Minimum Number of Channels With Hits in a Cluster 10000 20000 30000 40000 50000 60000 70000 Background Rate, (Hz) 0.2 0.4 0.6 0.8 1 SN Efficiency

ROC Curve, Varying Minimum Number of Channels With Hits in a Cluster ROC Curve, Varying Minimum Number of Channels With Hits in a Cluster

Single Marley Neutrino in a 2.246ms drift window.

Define trigger: Is there a cluster which passes all of these cuts? Background rate: Count number of triggers due to backgrounds. SN efficiency: Did we have a trigger from a cluster with > 2 Marley hits?

Pick a parameter to float and keep the

• thers loose.

After tuning:

13

After looking at each of these efficiency plots by eye to select the ‘best’ value of each variable.

Adjacent channel tolerance Minimum channels in a cluster Adjacent time separation Number of adjacent time hits. Cluster ADC Sum Efficiency Background Rate (Hz)

3 2 20 2 350 92% 18.0 3 2 50 3 500 83%

Back to the Data

14

Find more hits per event with a different hit finder? Fair number of events below 10MeV.

Summary

15

• Applying a preselection to the hits based on the hit primitives is

quite powerful and greatly cuts down the number of hits that would need to go through clustering.

• It is possible over 1000 events to get the background rate down to

zero at 83% single Marley neutrino efficiency for 5<Nu E<45MeV.

• At 92% efficiency background rate 18Hz for this ‘mini’ detector
• geometry. To high.
• Need more stats and to use the fast hit finder to investigate the

effect of hit ADC threshold.