Calorimeters offline calibration
PHOS & EMCAL
13 May 2009
Basics of Calorimeters Offline Calibration
• Calorimeters (PHOS and EMCAL) offline calibration requires the full
available statistics:
– PHOS needs ~109 - 1010 events estimated from a requirement to have 100-1000
0’s photons per cell,
– EMCAL might need less statistics due to a larger (7.5) cell.
• Calorimeters offline calibration is an iterative procedure, needs 5-10
iterations with the full statistics.
• Objects needed for calorimeter calibration:
–
–
–
–
–
ESDCaloClusters or AODCaloClusters (clusters energy and coordinate)
ESDCaloCells or AODCaloCells (cells energy and ID)
Vertex V0
Tracks (only needed for particular calibration algorithms)
Geometry from OCDB GRP
• Since the input for Calo calibration is ESD or AOD, the algorithm is
implemented as an Analysis Task
13 May 2009
Calo Offline Calibration
2/15
Time-dependent corrections to calibration
•
Since the calorimeters need many-months statistics to achieve required
calibration parameters accuracy, one has to take into account an evolution of
detector properties in time:
– Light yield vs temperature
– Gain variation with HV, radiation dose, rate effects
•
•
•
•
•
Online DA are supposed to collect information necessary to follow such
evolution of detector properties
LED runs (PHOS) or LED events (EMCAL) can be used to follow variations of the
gains in time
In general, calibration parameters can be expressed as c=c0+c1(t) with a
constant term c0 and a time-dependent term c1(t). c1(t) may be aggregated for
a group of channels (FEE cards, modules, supermodules).
Offline calibration can access OCDB to apply corrections to cell amplitudes
ESD QA can be useful to judge whether time-dependent correction should be
applied.
13 May 2009
Calo Offline Calibration
3
Each (i-th) iteration does:
1. Opens ESD/AOD, reads calibration parameter corrections calculated from
(i-1)th iteration
2. Writes the output of the Calibration Analysis Task as a set of N histograms
TH1 (N is a number of cells, N=18k for PHOS and N=11k for EMCAL)
3. Calibration analysis task analyzes data of several chunks in parallel (GRID
jobs). Output histograms from all chunks should be merged to create an
accumulated set of N histograms. Some bookkeeping is required to mark
the merged chunks, in order to avoid double merging and missed chunks.
4. As soon as enough statistics is accumulated, the histograms are analyzed
to calculate the new set of calibration parameters corrections
5. The output histograms and calibration parameters corrections are
transient objects:
– no need to store them in a persistent mass storage system.
– Should survive until the i-th iteration processes the full statistics and full merging is
completed
13 May 2009
Calo Offline Calibration
4
End of iterative calibration
• Criteria of the calibration procedure convergence:
– Variation of the calibration parameters corrections in the (i-1)th and i-th
iteration is less then 1;
– Variation of the 0 mass distribution width in the (i-1)th and i-th iteration is
less then 2.
• After the final iteration, a new set of calibration parameters is
calculated as a product of the initial calibration parameters
from OCDB and the calibration parameters corrections.
– Note that calculation of the new set of calibration parameters may
require manual intervention.
• The new set of calibration parameters should be submitted to
production managers to register to OCDB
13 May 2009
Calo Offline Calibration
5
Towards the offline calibration framework
by Cvetan & Chiara
Follow the presentation at the Weekly offline meeting 24 April 2009:
http://indico.cern.ch/conferenceDisplay.py?confId=57668
• In the first reconstruction pass, calorimeters can create their own AODs with PHOS or EMCAL
objects, and V0 (sometimes, Tracks). These AODs will be light-weighted and can be easily
handled by the calibration iterations.
• The first iteration can run in the first reconstruction pass, just after the reconstruction, with
the ESDs created by this reconstruction.
• ESD produced after the first reconstruction pass should be filtered to extract branches
CaloClusters, PHOSCells, EMCALCells. Some calibration algorithms using track matching
require the Tracks ESD branch which also should be extracted from ESD in the filtering
procedure.
• Usefulness of ESDfriends for Calo Calibration is obscured:
–
–
–
Output object merging is provided via TFileMerger.
But TFileMerges is not capable to merge thousands of files with 105 objects: an attempt to merge 1000 files
exhausted 16 Gb of virtual memory!
Output object merging is possible for pair of files: for example, output of each chunk should be merged with
accumulated output.
13 May 2009
Calo Offline Calibration
6
A step after offline calibration
• Besides creating the new OCDB object, we aim to recalibrate the ESD
created after the first reconstruction pass. For this, this first ESD
should pass through another filter, and parameters of the CaloClusters
(energies, coordinates, shower shape) should be recalculated with the
use of corrections to the cell amplitudes found by recalibration
procedure.
– As a result of this filtering, a new AOD will be created with better
calibrated CaloClusters. This new AOD should be used for the official
physics analysis.
13 May 2009
Calo Offline Calibration
7/15