EMCal DCS Status By B. S. Nilsen, M. Cherney, J. Fujita Creighton University For the EMCal project.
EMCal Offline Status
-
Upload
gillian-livingston -
Category
Documents
-
view
27 -
download
0
description
Transcript of EMCal Offline Status
05/07/2010 EMCal Status 1/29
EMCal Offline Status
Gustavo Conesa Balbastre
05/07/2010 EMCal Status2/29
EMCAL status @ Point 2 4 Super Modules installed
2 SM in C side 2 SM in A side
Trigger under commissioning. Bad channels
For LHC10b/c data there are 34 bad channels: 16 hot, (one FEC mainly) removed from reconstruction 3 dead 15 warm, just a flag to be followed in analysis.
For LHC10d the situation improves There is a T card in SM 1, becoming noisy from time to
time. About 100 channels are under investigation, MIP peak
was not observed / fitable, maybe just bad calibration.
05/07/2010 EMCal Status3/29
2009 vs 2010 runs 2009 runs
EMCAL: Clear 0 peak but misplaced at 105 MeV (instead of 135 MeV)
Main reason: Automatic pedestal subtraction.
Pedestal was calculated in the region where the signal starts to rise, therefore, the signal was biased.
2010 runs The position of the peak is now
reasonable ... We set fixed pedestals calculated
in pedestal runs. EMCAL
Only from run 115393 EMCal data can be used for analysis.
For previous runs, problem with sparse mode, part of the data lost.
p+p 900 GeV - pass 6
p+p 7 TeV - run 115393 - pass1
05/07/2010 EMCal Status4/29
Calibration With the accumulated statistics on LHC10b and
LHC10c we might have enough to attempt the calibration.
First attempt with MIP channel by channel has produced improvements that went to pass2.
We will try now calibration with 0 channel by channel.
Machinery in place in $ALICE_ROOT/PWG4/CaloCalib We have seen some electrons, but it will be hard
to calibrate with them. They can be good to test alignment and track matching, though.
05/07/2010 EMCal Status5/29
Calibration with MIP MIP peak cosmic
calibration set at 260 MeV
Data shows the peak close.
Use fitted mean per channel and equalize all.
05/07/2010 EMCal Status6/29
Calibration with MIP
Mean of MIP peak per channel. Use this to calculate the recalibration parameters
05/07/2010 EMCal Status7/29
Calibration with MIP, effect on 0
Both in pass1 and pass2, mass grows with momentum Non linearity effects are not corrected yet. We need still to improve calibration. In pass2 now mass is a bit higher
Mass resolution improves something. Still need more iterations to get to the expected
resolution Simulationp+p Min Bias 10 TeVEMCAL PPR
05/07/2010 EMCal Status8/29
Calibration with MIP, effect on 0 per supermodule
Improvement in Supermodules 1 and 2, the rest give similar resolution
05/07/2010 EMCal Status9/29
Track Matching
Standard matching cuts: |Δx|<6cm, |Δy|<6cm, |Δz|<6cm, ΔR<10cm Size of cell side is 6 cm Loosing the selection using pt dependent cuts represented by lines improves efficiency a 10 %, still need some improvement.
Task 2632: Track matching improvement Matching done during analysis, after reconstruction (this slide) seems
rather OK, although should be improved. Matching during reconstruction is much worse, should give the same,
under investigation.
05/07/2010 EMCal Status10/29
Track Matching: Electrons
Select macthed clusters with tracks looking to dE/dx:65<dE/dx<80 & R < 0.014
We are able to see electrons clearly.
0.5<pT<1.0 1.0<pT<2.0
2.0<pT<3.0 3.0<pT<5.0
10b pass1v4-19-15 AN
0.5<pT<1.0 1.0<pT<2.0
2.0<pT<3.0 3.0<pT<5.0
10b pass2v4-19-15 AN
05/07/2010 EMCal Status11/29
Track Matching: electrons residuals
Select electrons in 0.7<e/p<1.2 Residuals for e+e- in Φ,η as a function of momentum Eta residuals not centred at 0, effect of non corrected misalignment.
ΔηΔΦ
p (GeV/c) p (GeV/c)
-pos-ele
10b pass2v4-19-15 AN
05/07/2010 EMCal Status12/29
Geometry Alignment: Task 2533: Finalization of the detector geometry
checking for overlaps. Detector geometry implemented is ideal, no overlaps in this case. Space frame implementation is OK. Preliminary alignment corrections showed no overlaps, so this task is
closed.
Most urgent thing, put the final misalignment corrections in the OCDB
We are debugging the procedure to get the correction parameters calculated from survey data.
Some cleaning, rewriting of the geometry (mapping, position calculations) is maybe needed in near future ...
05/07/2010 EMCal Status13/29
Raw: fitting We have set a task force to set develop a fast fitter, that can be
used either in offline and HLT with good performance. Persons involved
David Silvermyr: Task coordinator and benchmarking. Per Thomas Hille: Peak finder fitter implementation. Paola La Roca: Neural Network fitter. Alexei Pavlinov: Fast Fitter.
Fitters currently in the code: Slow fitters (truth for benchmarking comparisons):
kStandard: default fitter used until now, Minuit fit, too slow. kLMS: refined version of kStandard
Fast Fitters kFastFit kNeuralNet kPeakFinder kCrude: Takes the maximum bin.
Need to do some benchmarking tests with all the fitters on the grid. We will do the request soon.
Before final decision: we want to implement small LUT table to handle small signals (95%) without any fitting.
05/07/2010 EMCal Status14/29
Recent developments in release: Raw
Signal thresholds: Threshold set to be amp > 3 units Fitted samples are float and could give 3.1 If unsuccessful fit, maximum taken, if 3, then it was
rejected. Changed for pass 2 threshold to amp >=3
Fit quality: The calculation of the raw signal fit Chi2 and NDOF can
give an idea of the quality of the amplitude fit obtained. Digits include these 2 parameters, but they are not
used, right now in reconstruction. Under study: Remove cells with large Chi2? How large?
05/07/2010 EMCal Status15/29
Recent developments in release: Digitization
Digits trimming: Electronics should give High Gain signal or High Gain plus Low
gain, but never Low Gain only. In some cases this was not the case After all the fits are done, such cases are removed from the
list of digits. Also those digits with abnormal times. Digits amplitude:
Until recent, digit contained an Int_t amplitude. We do fit the amplitude and get floats as result, we were
cutting our signal. Clear effect in spectrum. Now digits amplitude is float. This meant changes in many places in reconstruction and
simulation code.
05/07/2010 EMCal Status16/29
Effect of changes in digitization
Cluster energy distribution for different cuts on the number of cells in the cluster. Pass one shows repetitive structures that disappear with changes for pass2.
Run 117220pass1
Run 117220pass2
05/07/2010 EMCal Status17/29
Redefinition of digits and related tasks
Task 2538: Correct treatment of detector signal in sdigits correct event merging implementation A digit was an amplitude and time per channel This approach does not take into account in simulation:
Fitting problems of the different algorithms. Samples with different time arrival, pile-up.
How can we improve it: Produce a raw time sample from the energy lost and get the final amplitude from the fit of this sample.
All the placeholders for the samples (HG and LG) are since recently in AliEMCALDigit.
What needs to be done:1) Modify AliEMCALDigitizer::Digitize() so that it takes the Sdigit energy, converts
it in a time sample and then fits it. The sample and the fitted energy/time is kept in the digit.
2) In order to have a realistic performance, we need to implement a more realistic shaper : energy to time sample. The actual shaper is not realistic at low amplitudes.
Task 2537: Verification of event merging procedures Event merging of simulated data with the actual definition of Digits works.
Need to revisit when previous task is in place. Need to implement the embedding (merging of real data+simulated data).
Should we merge the time samples or the fitted values?
05/07/2010 EMCal Status18/29
Time resolution in simulation Task 2535: Implement realistic time resolution in digits.
In the code now we have a fixed time resolution of 0.6 ns We have to implement a resolution depending on the amplitude. Investigation in
parallel with raw fitting benchmarking.
To be implemented soon a correction based on this plot.
05/07/2010 EMCal Status19/29
Time measurement Task 2534: Handling of the time information from hits during digitization
The time passed to the digit looks reasonably good, so this task is closed.
We will need to revise it when we better understand the real measured time.
Very different value of the tower time measured in simulation and data. Need to shift the measured time
A constant shift due to our electronics, under study; and using T0 or TOF information.Simulation-LHC09d10
Run 117220 - Pass 2
05/07/2010 EMCal Status20/29
pass1
Pass1 vs Pass2, run 117220
pass2
Time cuts help to reduce noise, cut for pass 2 at 425 ns <t<825 nsTime signal seems to jump from one RCU to other, and even from run to run, under investigation
05/07/2010 EMCal Status21/29
Detector response Task 2627 : Correct detector response, GEANT and FLUKA.
GEANT3 simulations and beam test show a few % discrepancy.
A correction approach defined: Correct at the digit level, knowing the resolution measured in simulation and beam test.
Need to find a simple correction function. Wait for new beam-test results.
Fluka cannot be used now. What about GEANT4?
Eva has compared GEANT3 and GEANT4, there are differences. In order to correct any difference, we will have to use a different
sampling fraction: Need a way to know which MC was used, and this is not available.
Need to compare beam-test (old/new) data with simulation.
Photon Pi+
05/07/2010 EMCal Status22/29
Trigger Many changes in the code from last report:
Access to OCDB for simulation parameters and DCS information. Code still under development, many improvements ready to
be committed Mapping/numbering now obey ALICE convention, developed new
mapping methods to navigate between tower, FastOR & STU data
Decoding of trigger raw data Issue with EMCAL and dependence with VZERO solved. We
only use AliESDVZERO information. We want to change what is stored on ESDs for final analysis:
In ESDs: Trigger bit maps (L0, L1 (gamma), L1 (jet)). In ESDfriends; all complementary information to insure trigger
data is OK. Under study what would be the overhead. Old AliEMCALTrigger still kept since called by the actual
offline trigger frame.
05/07/2010 EMCal Status23/29
Reconstruction tasks Task 2633: Cluster unfolding implementation for !=0
Implemented the “simple” case = 0, off by default. Common task with HLT, Berkeley/Subatech groups interested in
this task. Need to have this working well before HI runs.
Need to test the code performance in PbPb simulations Retune reconstruction parameters Depends also on the unfolding.
Tasks 2630 : Study effect of misalignment in reconstruction. As said previously, need to produce misalignment object from
survey data, but still not available.
05/07/2010 EMCal Status24/29
Recent developments in release: Cluster Reconstruction
Cluster reconstruction at eta=0 borders EMCAL supermodules touch at eta=0. Until recent, no cluster could be shared by
both SM, which is nonsense. The sharing of clusters by 2 super modules
close to eta=0 implemented and used for pass2
05/07/2010 EMCal Status25/29
QA Reconstruction QA
Task 2551 and 2554: Implementation of run type and reference data
Task 2642: Extract QA data from MC production Tasks are partially done for pp data, not urgent tasks.
Analysis QA Class running on different trains
PWG4/PartCorrDep/AliAnaCalorimeterQA Produces many histograms. Under development a procedure to check from those
histograms, with simple markers any problem on run by run basis:
Average number of clusters, cells, and cells per cluster Average energy Average number of “pi0 pairs” Average time signal
05/07/2010 EMCal Status26/29
QA, pass1 Runs before 115393
(red circles) not usable, we see a clear jump.
Other jumps, related to aliroot, ocdb updates, need to concentrate on pass2.
Runs analyzed:114783,114786,114798,114918,114920,114924,114931,115186,115193,115310,115315,115318,115322,115393,115401,115521,115880.115881,115887,116079,116102,116118,116123,116134,116198,116287,116288,116402,116403,116562,116571,116572,116574,116642,116643,116645,116684,117034,117041,117048,117052,117054,117059,117063,117065,117077,117098,117099,117109,117112,117118,117220,117222.
05/07/2010 EMCal Status27/29
QA pass2, LHC10b Picture a bit more
stable, but there are some trends/jumps to understand.
When basic cuts (energy, single cell clusters removed) are applied, the picture looks a bit more stable.
Runs analyzed:115393,115401,115521,115880.115881,115887,116079,116102,116118,116123,116134,116198,116287,116288,116402,116403,116562,116571,116572,116574,116642,116643,116645,116684,117034,117041,117048,117052,117054,117059,117063,117065,117077,117098,117099,117109,117112,117118,117220,117222.
Cluster pairs with mass 110-160 MeV/c2
05/07/2010 EMCal Status28/29
TClonesArray handling / Memory Leaks
AliEMCALLoader owned several TClonesArrays. The EMCAL loader owned the data members with the list
of hits, digits and rec points in TClonesArray format. The loader is deleted often, so the same these lists. A loader is not supposed to own just to load. These datamembers have already being removed in
trunk. Besides, when dealing with hits, we were moving
the hits stored per track into a new common TClonesArray for all hits.
A correction is prepared and ready to commit. Under way, tracking down other leaks, bad
handling of TClonesArrays.
05/07/2010 EMCal Status29/29
Documentation-Changes log Place for the documentation in
http://aliceinfo.cern.ch/Offline/Detectors/EMCALOffline.html
A document for beginners also available http://aliceinfo.cern.ch/export/sites/AlicePortal/Offline/
Detectors/EMCAL/EMCal_Beginners.pdf It also includes a section with the ESD/AOD information
stored for the calorimeters. http://aliceinfo.cern.ch/Offline/Detectors/EMCAL/esd_aod.ht
ml Still a lot of things to complete.
Recent changes in the code/ocdb are reported in this wiki:
https://twiki.cern.ch/twiki/bin/view/ALICE/EMCalOffline