ProtoDUNE single phase noise Linda Cremonesi University College - - PowerPoint PPT Presentation

ProtoDUNE single phase noise

Linda Cremonesi

University College London

December 10, 2018

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 1 19

Introduction

I started looking at the noise in the protoDUNE single phase data The idea is to look at the noise in each frequency bin and for each channel, fitting Rayleigh distributions and identifying patterns as a function of time/channel So far I’m using np04 decoded-raw v07 06 00 allruns:

Run 4643 (DA: Physics! Tracks show 60 cm drift. Beam halo evident.) Run 5249 (DA: Physics. HV = 140. +1 GeV beam. First run after fixing trigger board to veto electrons.) Run 5571 (DA: CRT trigger test. N=4 gate (80 ns), all US-DS matches.) Run 5791 (DA: Physics. 180 kV. +3 GeV.)

I’m using gallery to get RawHits information and

libRootFffwWrapper for FFTs

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 2 19

What’s the idea behind this analysis?

1 Take the RawHits waveform coming from a specific channel:

500 1000 1500 2000 2500 3000

3

10 × Time [ns] 150 − 100 − 50 − 50 100 150 ADC

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 3 19

What’s the idea behind this analysis?

1 Take the RawHits waveform coming from a specific channel: 2 Take the Fourier transform:

0.0002 0.0004 0.0006 0.0008 0.001 Frequency [GHz]

2

10

3

10

4

10

5

10

6

10

7

10 Amplitude [ADC/GHz]

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 4 19

What’s the idea behind this analysis?

1 Take the RawHits waveform coming from a specific channel: 2 Take the Fourier transform:

0.0002 0.0004 0.0006 0.0008 0.001 Frequency [GHz]

2

10

3

10

4

10

5

10

6

10

7

10 Amplitude [ADC/GHz]

3 Take the FFT amplitude in each frequency bin and put it in a

histogram

4 Repeat for all events until we have one histogram per frequency bin

(> 3000 in total)

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 5 19

Rayleigh distribution

If the noise is purely thermal, the histograms can be fitted with a Rayleigh distribution: f(x) = Ax

σ2 · e

x2 2σ2

(1) where A is a normalisation constant and σ is the Rayleigh amplitude The Rayleigh amplitude, σ and the χ2 of the fit can tell us if the noise in the channel was thermal and how noisy it was

50 100 150 200 250

3

10 × Amplitude (ADC/MHz) 10 20 30 40 50 60 70 Events/bin ProtoDUNE average power spectrum Rayleigh Distribution for 0.17 MHz bin

Rayleigh fit results constant: 4.10e+06 +/- 0.00e+00 : 4.53e+04 +/- 7.41e+02 σ /NDF: 30.90 / 44

2

χ

20 40 60 80 100 120

6

10 × Amplitude (ADC/MHz) 5 10 15 20 25 30 35 Events/bin ProtoDUNE average power spectrum Rayleigh Distribution for 0.06 MHz bin

Rayleigh fit results constant: 1.96e+09 +/- 0.00e+00 : 5.60e+07 +/- 2.63e+06 σ /NDF: 67.91 / 25

2

χ

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 6 19

For a given channel in one subrun...

For a given channel in one subrun one can track all 3000 histograms, fit them all to Rayleigh distributions and in the end plot the χ2 or the Rayleigh amplitude σ as a function of frequency:

Frequency [MHz] 0.2 0.4 0.6 0.8 1 / NDF

2

χ 0.5 1 1.5 2 2.5 3 3.5 4 4.5

rChiSquares/rNdf:powX

0.2 0.4 0.6 0.8 1 Frequency [MHz] 20 40 60 80 100 120

3

10 × Amplitude [ADC/MHz]

Rayleigh Amplitude Rayleigh Amplitude

ProtoDUNE has 2560 channels per APA, and 6 APAs, resulting in 15360 total channels. Tracking 3000 histogram for every channel is not ideal, as it becomes > 46 million histograms ...

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 7 19

How do I deal with it?

Instead of tracking 15360 channels together, I now only track 100 channels (resulting in 300k histograms), and loop over the same subrun 154 times It’s quite time consuming (about 24 hour for 1 subrun) but I couldn’t think of any other way of doing it... The output is a tree with all Rayleigh parameters saved for each channel By default the histograms are not saved, if one wants to look at a particularly bad/interesting channel, a custom program can save all Rayleigh histograms for that channel In the next slides are some example plots that I can make from these Rayleigh trees

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 8 19

Compare Rayleigh Amplitudes of different runs: FEMB 0

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 9 19

Compare Rayleigh Amplitudes of different runs: FEMB 5

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 10 19

Compare Rayleigh Amplitudes of different runs: FEMB 9

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 11 19

Compare Rayleigh Amplitudes of different runs: FEMB 13

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 12 19

Compare Rayleigh Amplitudes of different runs: FEMB 17

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 13 19

Compare Rayleigh Amplitudes of different runs: FEMB 19

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 14 19

Bad channels

When reduced χ2 is higher than 10 for more than 1000 frequency bins → bad channel Comparing my bad channels with David Addams bad channel and sticky codes lists:

200 400 600 800 1000 1200 1400 1600 1800 2000 2200 Offline channel number 1 2 3 4 5 6 7

DA sticky DA bad run 5791 run 5571 run 5249 run 4643

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 15 19

Some initial look at phases

In our noise simulation we assume that phases are random Are they? I started look at the differential phase (ie phase in frequency bin n - phase in frequency bin n-1) in protodune channels If phases are purely random, the differential phase should average

• ut to 0

Next page some differential phases for some random channels in protodune runs 5571 and 5791

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 16 19

Differential phases

Run 5791, APA 0, some random channels

0.2 0.4 0.6 0.8 1 powX 0.06 − 0.04 − 0.02 − 0.02 0.04 0.06 0.08 avgDiffPhases

avgDiffPhases:powX {apaChannel==2}

0.2 0.4 0.6 0.8 1 Frequency [MHz] 0.05 − 0.05 0.1 0.15 Average Differential Phase

avgDiffPhases:powX

0.2 0.4 0.6 0.8 1 powX 0.05 − 0.05 0.1 0.15 0.2 0.25 0.3 0.35 avgDiffPhases

avgDiffPhases:powX {apaChannel==900}

0.2 0.4 0.6 0.8 1 powX 0.04 − 0.02 − 0.02 0.04 0.06 0.08 0.1 avgDiffPhases

avgDiffPhases:powX {apaChannel==1500}

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 17 19

Example histograms in specific frequency bins

hDiffPhases_6 Entries 9997 Mean 0.09437 Std Dev 1.7 3 − 2 − 1 − 1 2 3 Differential Phase 120 140 160 180 200 220 Entries hDiffPhases_6 Entries 9997 Mean 0.09437 Std Dev 1.7

Differential Phases: APA 0 chan 150, freq bin7

hDiffPhases_312 Entries 9997 Mean 0.09979 Std Dev 1.81 3 − 2 − 1 − 1 2 3 Differential Phase 130 140 150 160 170 180 190 200 210 Entries hDiffPhases_312 Entries 9997 Mean 0.09979 Std Dev 1.81

Differential Phases: APA 0 chan 150 ifreq 312

Plots are 0 suppressed Clear peak above 0 for frequency bin 7, but it looks like there’s a flat population too

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 18 19

Summary and next steps

Started comparing Rayleigh distributions between different protodune runs: normalisation difference → investigating Periodic resonances in some frequency bins? Started looking at differential phases, some channels show deviations from 0 at low frequencies → currently investigating patterns, looking at fembs, planes, inside/outside facing..

L.Cremonesi (UCL) ProtoDUNE noise December 10, 2018 19 19