Upload
dale-bradford
View
218
Download
0
Embed Size (px)
DESCRIPTION
3 Module-by-module ADC->keV Fit the p-p dE/dx distributions for each SDD module with a Landau+Gauss 85% of modules have ADC value within 5% from expected value AliITSresponseSDD modified to contain 260 SDC->keV conversion factors (1 per module) Committed on Jan 13 th 2010 Improved calibration used since pass4 Layer 3Layer 4 with E. Biolcati
Citation preview
1
SDD offline statusSDD offline status
Francesco PrinoINFN sezione di Torino
ALICE offline week – March 15th 2010
2
Summary of main changes since Summary of main changes since October 2009October 2009
Module-by-module ADC->keV calibration Modified OCDB objects, calibration constants from p-p 2009 data
Updates in noise rejection in Cluster Finder Improved efficiency for large drift times
Improvements in Injector DA Avoid artifacts of the polynomial fit to drift speed vs. anode
Fixes for signal embedding Few updates needed in the treatment of raw->SDigits conversion
Improved treatment of anode gain in simulationAlignment Millepede results for Alignment+Calibration Ongoing cross-checks
QA + Pilot Train
3
Module-by-module ADC->keVModule-by-module ADC->keVFit the p-p dE/dx distributions for each SDD module with a Landau+Gauss 85% of modules have
ADC value within 5% from expected value
AliITSresponseSDD modified to contain 260 SDC->keV conversion factors (1 per module) Committed on Jan
13th 2010 Improved calibration
used since pass4
Layer 3 Layer 4
with E. Biolcati
4
Module-by-module ADC->keVModule-by-module ADC->keVFrom pass4 with module-by-module ADC calibartion
with E. Biolcati
Cluster selectionCluster selectionLoss of efficiency observed with increasing drift time in pass2 data Larger loss on layer 3 than on layer 4
Reason for the efficiency loss: SDD clusters removed by the cut on the cluster size applied to remove noise clusters Cluster size increases with increasing drift time due to charge diffusion
The larger loss on layer 3 possibly due to the higher temperature which causes larger charge diffusion Cuts on cluster
size released to recover missing efficiency Committed to
trunk on Feb 3rd 2010
OK in pass 4 reconstruction
Region close to anodes (small drift times, latge |xloc| under investigation
5
Run 104892 pass 2 Run 104892 pass 4
with A. Dainese
SDD efficiencySDD efficiencyEffect of the looser cuts on cluster size which allow to recover in pass4 the missing efficiency at large drift times + new objects in OCDB sim (closer to real detector) for simulation Few modules still show discrepancy between data and MC under
investigation.
6
Run 104892 pass 2
Run 104892 pass 4
Dat
a/M
C
with A. Dainese
7
Injector DAInjector DADrift speed extracted online from injector runsUse a polynomial function to model the vdrift dependence on anode coordinate Need a 3rd degree polynomial Artifacts in the fit for some modules due to missing injectors
on the bordersImproved fit imposing these conditions: Vdrift maximum close to the center of the sensor No other points with derivative = 0
All injector runs from p-p data taking have been reprocessed and improved DriftSpeedSDD objects have been put in OCDB, just before pass5 Allowed also to improve the Millepede alignment
8
Raw data to SDigitsRaw data to SDigitsChecking of embedding for ITS showed anomalous results for SDD (M. Poghosyan)Problems due to the fact that the raw->Sdigits conversion method was not updated since long time and did not contain the upgrades applied in the RawStreamSDD decoder in the last year Crash for some events
Due to missing decoding of newly introduced words Anomalous results: missing clusters in the merged event,
wrong cluster coordinates Due to “old-style” treatment of zero suppression threshold and drift side
All problems (mainly in AliITS.cxx) fixed with these two commits: 38956 (Feb 17th) and 39464 (Mar 9th)
9
Treatment of gain in simulationTreatment of gain in simulationGain is measured anode-by-anode in PULSER runs and stored in the OCDB in the AliITSCalibrationSDD objects (in CalibSDD) NOTE: CalibSDD used iboth in simulation and reconstruction
In RECONSTRUCTION: the anode gain is “normalized” to the average gain before being used to correct the measured ADC counts Gain correction is meant to equalize the gain among different anodes,
the “absolute” calibration is done with the ADC->keV conversion factorIn SIMULATION the anode gain was applied as it was to the ADC counts Due to the fact that in the ICDB sim the simulated gain had average = 1
by construction, so the normalization to the average gain was not needed
Weak point gave rise to miscalibrated signals when “real” detector parameters were put in OCDB sim
Pass4 simulations had dE/dx off by ≈30% due to this problem (affects only SIM, not DATA) Not a blocking problem: dE/dx can be easily renormalized at the analysis level
Improved treatment of gain in simulation (introducing the normalizaation to the average gain) committed on Feb. 25th
ADC->keV conversion factors in OCDB sim updated accordingly
AlignmentAlignmentSpecial feature for SDD in Millepede II: calibration parameters (TimeZero and drift speed correction) added on top of geometrical misalignments as free parameters in the fit to track-to-point residuals Time Zero initial values estimated from residuals in the two drift
regions , in the future better to use minimum drift time Vdrift correction needed for:
Modules with malfunction injectors ( ≈ 30%) Systematic effects in the estimation of the drift speed with injector
10
R. Shahoian
Alignment: checks (I)Alignment: checks (I)Fix modules w.r.t. ladders to the survey measurements and leave free in Millepede: TimeZero and vdrift for each module Rotations and translation for ladders
11
R. Shahoian
Alignment: checks (II)Alignment: checks (II)Use run 104892 to extract parameters and compute residuals on run 104321 Run 104321 is “served” by a different DriftSpeedSDD
object, so it can be used to check if the drift speed corrections extracted from one set of runs can be applied to other runs
12
R. Shahoian
Black = only surveyRed = Millepede alignment+ calib
Residuals (cm) Residuals (cm)
Alignment: on-going and to-doAlignment: on-going and to-doDrift time distribution at low times under investigation Peak close to time zero
Due to particles crossingthe sensors beyond the anodes
Loss of efficiencyMay be due to online zero
suppression, being checked
Try to fix TimeZero for half ladders in MillepedeIIValidate correction maps and include them in reco For about 10% of the modules significant non-
uniformities of the drift field are present and have been mapped with laser measurements during construction
Should improve significantly the track-to-point residuals13
14
Correction mapsCorrection mapsResiduals along drift axis vs. drift coord. from p-p data after Millepede with and without correction based on laser maps Tested on 5 modules with larger effects
GOAL: Validate the correction extracted from laser Promising results, but more checks are needed
Xloc
Res
idua
l (cm
)Xl
oc R
esid
ual (
cm)
Xloc (cm)
Millepede
Millepede + Laser map
S. Beolè
15
QA/DQMQA/DQMAliITSQASDDDataMakerRec: ResetDetector added: reset of the histograms and set to
zero of the calibration objects Raws and in the RecPoints lists: added the normalized (to
the number of alive channels taken from OCDB) histograms of the module pattern and of the layer patterns. Histograms set as Expert histograms.
AliITSQASDDChecker: Raw Data: simple checks RecPoints: simple checks (not in the trunk)
AliITSQASDDDataMakerSim: All histograms have been implemented
AliITSQASDDChecker: Check for SIM histos To be implemented
M. Siciliano
16
SDD Task in Pilot TrainSDD Task in Pilot TrainTask PWG1/ITS/AliAnalysisTaskSDDRP Included in Pilot Train to check SDD performance in first p-p
data Based on RecPoint+ESD+ESDfriend information Allow to do more detailed checks of performance and
calibration with respect to standard QA E.g. cluster charge corrected
for track inclination vs. drifttime
Will be improved (with new plots concerning efficiency and extra clusters) in the next days
17
Other ongoing activitiesOther ongoing activitiesScan of raw data to understand the problem of missing efficiency at low drift times Likely to be due to the zero suppression algorithm, but
needs some further investigationCluster selection in the cluster finder Develop more refined cuts (not only cluster size) to reject
noise clusters Develop a special treatment for clusters on the border or
close to a dead anodeSimulation Checks for the ≈10 modules that have not the same
efficiency in data and MC