Calorimeter Monitoring at DØ

Transcription

1 Calorimeter Monitoring at DØ Calorimeter Monitoring at DØ Robert Kehoe ATLAS Calibration Mtg. December 1, 2004 Southern Methodist University Department of Physics Detector and Electronics Monitoring Levels Data Quality Physics Monitoring Calibration and Resolution Final Comments

2 End Calorimeters (EC) Electromagnetic (EM) D0 Detector D0 Detector Fine hadronic (FH) calorimetry U LAr hermetic h < k cells 50 dead Dh x Df = 0.1 x x 0.05 (shwr max) 4 EM, 4-5 HAD depths all new analog electronics Coarse hadronic (CH) Central Calorimeter (CC) new preshower detectors designed to recover e/g E resolution lost to solenoid material new Si microstrip, fibers trackers 2 T solenoid inter-cryostat scintillato

3 Electronics drift time 450 ns changes since Run I bunch crossings 396 ns higher # mult. interactions per crossing shape signals to peak ~320 ns baseline subtraction (BLS) electronics sample and store (SCAs) ea. 132 ns retrieve on Level 1 trigger accept baseline subtract (396 ns earlier) pile-up and low freq. noise move to secondary SCAs on Level 2 accept, send to digitization stage outside of collision hall pulser Calorimeter Preamp/ Driver Trig. sum Filter/ Shaper x1 x8 alternative readout Bank 0 SCA (48 deep) SCA (48 deep) SCA (48 deep) BLS L2 SCA Output Buffer two gains for better dynamic range SCA (48 deep) Bank 1

4 Pulsers inject known charge into preamp Do this separately for gains x8 and x1, optionally also separately for the two L1 SCAs per channel. ADC (readout) debugging aims Identify technical problems in the electronics, like e.g. dead channels. calibration aims channel-by-channel differences in electronics response. non-trivial to model signal reflections from detector capacitance variations in cables, detector cells model each preamp and shaper type gives handle on non-linearities from electronics mainly from analog buffers (SCAs) DAC (pulser signal)

5 Online Monitoring Online Monitoring trigger/daq monitoring performed in trigger system Detector or physics specific monitoring each detector system has own monitoring processes full event assembled, post-trigger ~10% of recorded events used ROOT-based histogramming and ntupling software convenient browser for efficient investigation of potential problems stream 4 different streams (different types of triggers) different types of problems affect different triggers continuously monitored by shifters cell occupancies, jet and Etmiss districutions global monitor monitor data online using L1, L2 and L3 objects jets, electrons, muons allows to find problems not visible with detector quantities closer to what will affect analysis

6

7 Offline Monitoring Offline Monitoring use reconstructed objects jets, Etmiss, muons, electrons results available 2-3 days after data taken use reference histograms helps understand impact on physics separate monitoring of data quality via Etmiss measurement consider mean and RMS in x and y to diagnose problems

8 Data Quality Data Quality strict and systematic data quality control running in control room during data taking hot or missing cells, towers, BLS one algorithm compares cell to its neighbor coherent noise pedestal drift other hardware failures detected automatically in real-time: quick response time a good data quality strategy, including monitoring, assures quality of physics results event quality flag cell occupancy and RMS trigger and full readout comparison

9 Treatment if problem appears in run channel(s) can be masked --> 5-15 min. beam time lost this run affected; cell(s) tower(s) are lost if need calibration end of store is hours away if need hardware fix, request access, few days offline correction problems found offline and affected many runs go back and reprocess in some way to improve data quality not all problems correctible need good run/lumi-block selection hot cells that couldn t be fixed immediately are recorded in a database for offline killing

10 Readout Consistency Checks Readout Consistency Checks Several problems in early data found thru comparison of independent trigger and full calorimeter readout different effects on precision and trigger chains use Level 1 to correct or reject events coherent noise via ground loops very brief bursts in early Run largely eliminated thru extensive grounding tests precision readout early running: very rare now timing problems with firmware eg. wrong SCA cells read out hard to find in data with normal diagnostics pulsers timing slightly different than normal data trigger readout continues to play a key role very powerful for detecting non-trivial problems trigger readout usually not affected by problem can sometimes be used to determine correction Level 1 readout

11 Monitoring in Physics Analysis Monitoring in Physics Analysis problems found offline very costly: many runs can be affected need offline corrections for collected data offline correction subtle problems found offline and affected many runs go back and reprocess in some way to improve data quality trigger readout plays key role alternative readout usually not affected by same problem so can be used to determine correction problems observed in high Pt electron data problems in BLS readout electronics tails of MEt and jet Pt are extremely sensitive to data quality

12 Electronics glitches Some problems in our very early data were found in offline analysis of physics data. This plot illustrates such a problem. FIX ED Density of high-pt (> 25 GeV) electron candidates in a relatively small corner of one of the end-caps. Observed holes in a distinctive pattern. Traced down to a timing problem in the firmware, fixed in next firmware version. Due to this timing problem, the wrong gain was reported, mostly for events with very high energy depositions in single cells. Relatively hard to find, as only visible in rare events with very high single-cell energy deposits. Relatively hard to find using pulsers, as timing works slightly differently for pulser and collision data.

13 Some Notes Some Notes calibration model a guide for online monitoring test beam + pulser gives calibration but, more work always still needed: in situ has been done well after beginning of run should be part of online monitoring/characterization strategy What is not monitored, but should be Z -> ee for electron scale, efficiency and uniformity studies Z/photon+jet for jet scale, efficiency, uniformity and resolution studies these or other events for control samples to understand MEt resolution expedite physics commissioning maintain after establish initial state of things

14 Expected Electron and Etmiss Studies Expected Electron and Etmiss Studies relative scale in phi could now do at ATLAS relative scale vs. eta issues of cracks, dead material absolute scale per cryostat linearity using E/p Etmiss resolution and tails <E T /p T > DØ Collaboration, PRD 58, (1998) Fiducial region that was used in Run I precision measurements. Z Æ e + e - at least one electron in CC mod(f) [position in f relative to module boundaries]

15 Expected Jet Studies Expected Jet Studies absolute scale non-linearities and dead material should be model-able based on test beam need to establish it relative scales may be harder to model due to difficult transition regions very high statistics dijet and direct photon samples establish eta and phi dependence of jet scale to high precision

16 Final Comments Final Comments tails of MEt and jet Pt are extremely sensitive to data quality to study new physics, good data quality strategy should be applied trigger vs. full readout comparisons is very useful to detect, and even fix, problems prompt hardware validation extremely important online monitoring with hardware AND physics level diagnostics detailed tracking of analog electronics, radiation and other trends automation compensates for non-expert shift crew documentation, training, databasing understand detector fully as soon as possible calibration, resolution, stability in situ express stream Z->ee, mm samples, high Pt g samples Note: express streaming data to get interesting events quick is useless unless understand detector be ready with MC tuning work with first data