Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 1
HLT & Calibration
Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 2
HLT as Calibration Data Source– Among the main tasks of HLT (especially in the first
years)● Monitoring of the detector performance● Analysing calibration data (real events and calibration triggers)● Calculation of a (first) set of calibration parameters● Applying calibration to online event reconstruction● Storing parameters in relevant databases
– Procedures are being developed in collaboration with TPC, TRD, PHOS, and Dimuon detector
– HLT is open for calibration tasks for all participating detectors
● Compatibility w. Computing Power / Financial Budget
Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 3
HLT & Databases● The HLT will have contact to different databases, which are for
different purposes as well as under different administration.
● For now, it is foreseen to have contact to three different DBs:
– Configuration DB (DCS)
– Condition DB (DCS)
– Calibration DB (Detector Specific)
Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 4
HLT DB Remarks
• Use of dedicated HLT portal nodes to contact DBs
• Redundancy of these nodes to avoid single-point-of-failures
• Data of the DBs is cached (where possible) and distributed to the appropriate places inside the HLT cluster (using its publisher-subscriber framework)
• Output data to DBs will be written from the global layer of the HLT cluster.
• Contact to the databases will be performed via ODBC or similar techniques.
Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 5
TPC Calibration Example (1)– Pedestal run analysis and download– Pulser runs
● Accumulation of pulser events● Gain calibration and timing alignment● Map hot/dead pads
– Laser runs● Online reconstruction of laser tracks● Drift velocity map● ExB map● static E-field distortion map
– Electron attachment / gas contamination ● Online reconstruction of particle tracks / laser tracks● Cluster charge loss per drift length
Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 6
TPC Calibration Example (2)
–Drift velocity map» Nuclear collisions
» Accumulate charge per timebin
» Calculate charge step at central electrode
» E.g. NA49 data
NA49
charge step
Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 7
TPC Calibration Example (3)
–Space charge effects» Online reconstruction of tracks
» Histogram of track impact parameters
» E.g. STAR TPC – fluctuations on second timescale
Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 8
TPC Calibration Example (4)
– Histogram of track residuals» E.g. STAR TPC – grid leak distortions
• Ion leakage at gating grid gap
Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 9
PHOS Calibration Example– Pedestal run analysis and download
–LED runs» Accumulation of pulser events
» Gain calibration and timing alignment
» Map hot/dead channels
–Gain/energy calibration» Nuclear collisions
» Online pulse shape analysis
» Timing alignment (using fastest particles)
» Online reconstruction and accumulating of showers
» Relative gain calibration by equalizing channels
» Absolute energy calibration by monitoring pi0-peak
Sebastian Bablok – University Bergen -- Timm M. Steinbeck - University Heidelberg 10
HLT Questions/Open Points
To plan connections from HLT to DBs, HLT needs to know:
• At which places (processing steps) in the HLT cluster is DB data needed?
• What is the estimated processing budget for online calibration?
• What data for or/and from which databases?
• Which detector needs what data?
• How large is which data?
• (different for different data -> detailed list wanted)
• How often is data updated? (once (caching) regular requests to DB)
• Output data
• General
• e.g. Drift Velocity: • Type of measurement (once, continuous spectrum, every X sec., average, ...)
• Write to DB (continously, at end of run, ...)
• Who has access to data (online, offline)?