Timing Measurement in the CALICE Analogue Hadronic Calorimeter.

Transcription

1 Timing Measurement in the CALICE Analogue Hadronic Calorimeter. AHCAL Main Meeting Motivation SPS CERN Testbeam setup Timing Calibration Results and Conclusion Eldwan Brianne Hamburg 16/12/16

2 Motivation for precision timing Timing may help: Resolving components in hadronic showers -> ~ ns Timing based clustering/energy reconstruction -> ~ ns Background rejection and pile-up mitigation -> ~ ns Software compensation like Space-Time Development of Electromagnetic and Hadronic Showers and Perspectives for Novel Calorimetric Techniques IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 63, NO. 2, APRIL 216 ex: ɣ ɣ -> had background Hit time of pair bkg in vertex detector (SiD) arxiv: v1 2

3 What time resolution is needed? Contributions to time resolution (AHCAL specific): SiPM ~ - ps rise time (dependant of number of p.e) Scintillator ~ ps rising time Light propagation to the photodetector ~ ps to ns Electronics -> ps to ns Resolution needed is related to the physics: EM showers: quasi-instantaneous, O(ns) Hadron showers 1) Prompt component due to π /η decays + direct ionisation. O(ns) 2) Slow component due to nuclear excitations, recoil and neutron evaporation/scattering O( ns) CALICE T3B arxiv: v2 Time resolution needed ~ 1 ns from physics 3

4 Testbeam at CERN SPS Testbeam campaign at CERN in July 215 Goals: Test of different tile/sipm designs Check EM performance of the detector Study of timing of hadronic showers in 3D (radial, longitudinal) Setup: 14 layers (~ 38 channels) Trigger signal (T) directly fed to the chip as a normal channel reference time Trigger signal ~1m ~ 1.5 λ / 15 X * * Iron : λ = 16.8 cm, X = 1.7 cm Picture of AHCAL in steel CERN λ: interaction length X: radiation length 4

5 The readout chip : SPIROC2b Self-triggered chip / 36 channels Only the first hit is registered in a clock period Energy measurement : Analog measurement / 12 bits 16 memory cells per channel Time measurement : ADC sampling of 2 voltage ramps 16 memory cells per channel Testbeam (25 khz BX) : ~ 1.6 ns per bin Design to run synchronous to the ILC machine (5 MHz BX) -> time resolution ~ x2 better 2 t [ns] (ILC) 4 t [µs] (TB) 5

6 Calibration Overview slope per Chip, BXID pedestal per Chip, Chn, Mem = 1 except T channels 12 parameters (poly 2nd order) Raw TDC Time (ns) T correction and selection Shift to offset per Chip, Chn, Mem, BXID Ramp linearity Correction Time Walk Correction TDC nhits correction Final time of first hit distribution 3 parameters per Chip, BXID (poly 2nd order) 3 parameters (global) (expo + offset) 2 parameters (global) (linear function) Check time of first hit distribution Check time of first hit distribution Check time of first hit distribution ~ 2 constants! 6

7 Conversion from TDC to nanoseconds Timing is measured via a voltage ramp and stored in a memory cell Assumed to be linear Not possible to measure directly the ramp (too many chips) -> developing a robust procedure to extract the slope of the ramp Need to use instantaneous particles -> muons for calibration, electrons for cross-check Procedure: Measuring starting point and end point of the ramp on the TDC Spectrum Total ramp length in TB (25 khz) ~ 4 ns (392 ns due to dead-time) Extract slope for each chip, BXID Extract pedestal for each chip, channel and first memory cell In the order of expected value: ~1.6 ns/tdc slope chip,bxid [ns/tdc] = 392 ns Max chip,bxid Pedestal chip,bxid Events / TDC Entries /.5 ns/tdc Pedestal Mean 1.56 ns/tdc, RMS.12 (even BXID) Mean 1.56 ns/tdc, RMS.11 (odd BXID) Maximum Time [TDC] t[ns]=(tdc Pedestal chip,chn,mem=1 ) slope chip,bxid slope [ns/tdc] 7

8 Calibration of the time reference The time of the trigger is needed in order to have the relative time of a particle -> need proper calibration 4 channels in the AHCAL receiving the trigger signal from scintillators The trigger signal is coming from the exact same source Time of the trigger should be the same between channels. Calibration needed due to different pedestal values per ramp BXID for each memory cells Trigger resolution ~ 4 ns Electronics contribution in final resolution [ns] 14 T [ns] 14 - T 12 T # Events With Correction: Mean.9 RMS 4.82 No Correction: Mean.56 RMS T 12 - T 14 [ns] Time reference correlation after calibration T [ns] 8

9 Time of first hit distribution Calibration applied similarly to all channels Time of first hit is determined by looking at the time of the hit relative to the time of the trigger Slightly asymmetric distribution Time resolution ~ 5.7 ns RMS Possible improvements due to effects known from the electronics Non-linearity of the voltage ramp Time-walk effect due to the threshold # Entries / ns Resolution [ns] Muons (Simple Calibration) Mean σ = 5.66 ns Mean RMS = 5.82 ns Simple Calibration Muons Mean -.6 ns RMS = 5.65 ns Layer 9

10 Ramp non-linearity correction Assumption of a linear voltage ramp would mean: Time of the hit versus the TDC value of hit would yield a flat distribution -> no dependency of where the hit occurred in the ramp No totally correct -> small kink in the middle of the ramp Correction with a 2nd order polynomial done for each chip and ramp BXID (odd/even) Improvement in the order of 5.3% (5.35 ns RMS) # Entries / ns Module9 / Chip146 / BXID χ2/ndf: 71.45/56 p: ±.666 p1:.5541 ±.711 p2: e-6 ± 1.82e Hit Time [TDC] Non-Linearity Calibration Muons Mean -. ns RMS = 5.35 ns

11 Time-Walk correction Threshold -> dependency with hit energy Low amplitude hits trigger later than high amplitude hits Fit with a exponential function and assumed to be global Corrections up to ~ 6 ns between low energy hits and high energy hits Improvement in the order of 3% (5.19 ns RMS) χ2/ndf 299.8/14 a.9 ±.7 ns -1 b ±.1 MIP c ±.1 ns Hit Energy [MIPs] # Entries / ns All Corrections Muons Mean:.1, RMS: ns

12 Number of hits correction (electrons) Expected offset compared to muons (different trigger setup) ~ ns When several channels of a single chip are saturated -> pedestal shift Similar effect for the TDC is observed -> increase of the time of hit with the number of hits in a chip Increase of the time resolution up to -12 ns for large number of hits -> relevant for electrons and hadron showers # Entries / ns GeV e Max: -.5 ns RMS: 8.92 ns RMS [ns] offset + slope*x + A*exp(B*x) χ2/ndf 66.22/12 slope 1.71 ±.29 ns offset ±.1 ns A 1.6e-7 ± 2.6e-7 ns 8 B 1.18 ± Number of triggered channels over.5 MIPs Number of triggered channels above.5 MIPs 12

13 Tuning of the simulation Muons Resolution ~ 5 ns RMS Fit with double gauss function -> empirical, take into account asymmetry Use parameters extracted from data for simulation ~% max deviation over [-2, 2] ns range normalised entries MC/Data Muons Data Mokka (QGSP_BERT) DD4hep (QGSP_BERT)

14 Tuning of the simulation Electrons Resolution ~ 7.8 ns RMS Additional gaussian smearing parametrised from data Agreement within % Looking other energies: nice agreement for the data, MC is worse -> investigations normalised entries GeV e- 15 GeV e- 2 GeV e- 3 GeV e- normalised entries MC/Data GeV e- Data Mokka (QGSP_BERT_HP) DD4hep (QGSP_BERT_HP)

15 Conclusion and Outlook A procedure to calibrate timing in the has been developed Timing resolution in testbeam with SPIROC2b achieved between 5 to 7.8 ns -> should be enough to study timing in hadronic showers Outlook: Finalising electron data Look into pion data: time dependance in 3D (radial and longitudinal evolution) Short/long term correlations between layers for hadronic showers 15

16 Backup 16

17 SiPM used in Technological prototype Old ITEP tiles with WLS fiber 8 px New ITEP tiles Ketek 12k px EBU strip Hamamatsu MPPC k px - 36 px NIU megatile SMD MPPC 16 px (5 µm) MPPC 25 µm UHH + Heidelberg wrapped tiles Ketek 23 px SenSL 13 px 17

18 SPIROC Acquisition 2~5 us 4 us ILC mode idle initialize TB mode idle initialize BXID BXID 1 BXID 2 Channel 1 Channel External trigger (TB only) "keep data" (TB only, simplified) (missed in TB) (validation) (noise) (suppressed in TB) SPIROC TDC ramps (TB) Channel 1 TDC (TB) Channel 36 TDC (TB) SPIROC TDC ramps (ILC) Channel 1 TDC (ILC) Channel 36 TDC (ILC) 18

19 Muon Selection MIP Preselection nhits < 2, < CoGZ < 8 mm εmuon = 99.4%, εelectron <.1 % and εpions = 13.3% Track Selection: Track length: SSF (8), BL (3) Number of hits in a layer <= 2 εmuon 72.5%, εelectron <.1% and εpions 5.6% nhits 14 AHCAL Simulation Muons Electrons Pions MIP Preselection AHCAL CoG Z [mm] Fraction of Towers AHCAL Simulation Muons (Inner 12 12) Muons (Outer BL) Electrons Pions Fraction of Events AHCAL Simulation Muons (Inner 12 12) Muons (Outer BL) Electrons Pions Maximum number of hits in a layer Number of hits in Track 19

20 Electron selection Event Quality Cherenkov (Data only), Energy in 3 first AHCAL layers > MIPs Fiducial cut (9x9x25 mm 3 ), E13+14 < 1% Esum, 4 < nhits < (2 GeV) Efficiency: εmuon <.1%, εelectron < 87.5% and εpions = 5.1% nhits 14 AHCAL Simulation Muons Electrons Pions # Events AHCAL Simulation Muons Electrons Pions # Events AHCAL Simulation AHCAL CoG Z [mm] Muons Electrons Pions (E +E 4 +E 5 ) [MIP] Eldwan Brianne Timing Measurement in the CALICE Analogue Hadronic Calorimeter page (E +E 14 )/ E [%] 2

21 Layer 11 - Electronics problem Layer 11 shows significant worse time resolution Due to noisy TDC ramp -> double ramp present for all chips Difficulties to extract the ramp slope Worsen the time resolution significantly Events / TDC 3 2 Entries Additional ramp Time [TDC] 21

22 Layer 4 and 5 - Electronic problem Layer 4 and 5 (new ITEP) present a strange feature Double peak structure All chips and channels Related to shower maximum? # Entries / ns AHCAL e- 2 GeV Chip 161 after calibration Chip 161 after calibration (cut ntriggered < 15) Chip 161 after calibration (cut ntriggered above.5 MIP < 15) 2 #Entries / ns GeV e Module 4 Mean:.271 ns RMS: ns

23 Correction for trigger delay An offset is extracted to account for the delay of the trigger Trigger logic and cabling Offset extracted for each chip, channel, memory cell and BXID # Entries / ns 3 2 (Muons) First Fit Iteration Offset per BXID needed! Pedestal per memory cells different for each ramp BXID Offset BXID odd [ns] Offset [ns] Offset BXID even [ns] 23

24 Cross-check Cross-check considering single hits in a chip Time resolution should be similar to muons Ok! Time resolution very similar to muons Parametrisation of the RMS increase with number of hits Assumes a simple gaussian widening Implementation in simulation normalised entries GeV e All hits - RMS: ns Single hit - RMS: 5.49 ns Muons - RMS: 5.13 ns RMS effect [ns] 8 6 χ2/ndf 29.97/ 4 Constant 8.75 ±.3 ns Factor 1.4 ± Number of triggered channels over.5 MIPs 24

25 Asymmetry in muons Probably due to the non-linearity from the time reference Non-corrected for as no external reference Dependance visible function of the TDC value of the time reference Late hits seems to explain the asymmetry observed -> time reference more off along the ramp (lever arm) # Entries / ns Layer 3 < TDC T < < TDC T < 2 2 < TDC T < < TDC T < 3 3 < TDC T <

26 Influence of ROC Digitisation Checking influence of the ROC Digitiser on the timing As expected for muons, no change with threshold Electrons -> same conclusion # Entries / ns 12 AHCAL Muons ROC.5 MIP ROC.2 MIP ROC.3 MIP ROC.4 MIP ROC.7 MIP # Entries / ns.6.5 AHCAL Electrons 2 GeV ROC.5 MIP ROC.2 MIP ROC.3 MIP ROC.4 MIP ROC.7 MIP