CSC Detector Woochun Park @USC ATLAS Meeting Jan 3, 2007 Cathode - PowerPoint PPT Presentation

CSC Detector Woochun Park @USC ATLAS Meeting Jan 3, 2007

Cathode Strip Chambers • MDT are well suited in barrel region with economically produced chambers for the 5500m 2 area. • But, large diameter and high operating pressure make them unsuitable in areas where high counting rates are expected (>200Hz/cm 2 ). • In |h| > 2.0, CSC detectors are used. – It provides the required spatial resolution of 80mm. In several prototypes, a sigma <= 60 mm has been measured. – Electron drift time less than 30ns resulting an r.m.s timing resolution of 7ns. By detecting the earlist arrival from four layers, r.m.s. resolutions of 3.5ns have been measured. – Low neutron sensitivity: because of the small gas volume and the absence of hydrogen in the operating gas (Ar/CO 2 /CF 4 mixture), the measured sensitivity is less than 10 -4 . • Details are muon TDR. http://atlasinfo.cern.ch/Atlas/GROUPS/MUON/TDR/pdf_final/CSC. pdf

CSC Reconstruction Methods • Center of Gravity – This method uses the three amplitudes of the cluster to compute the centroid of the charge distribution. – Currently, it’s adopted in Moore package by David Adams. • Parabola Interpolation – This method draws a parabola through the three amplitudes and uses the position of its maximum to calculate the hit position. • Cluster Fit – This reconstruction method uses a 2-dimensional fit to the time samples of the center strip and its 2 neighbors on each side.

CSC Cluster Performance • The code is written and run in athena framework. • 12.0.3 codes are checked out. • atlas-dc3- 07.007233.digit_newtags.muminus_pt100GeV._00001.pool.root MC (20k) are used (in castor [CERN Advanced STORage System]) – DetDescrVersion = “ATLAS-DC3-07” should be used. – If not, ntuple doesn’t save any information. • CscSimPosValidatorOptions.py makes csc_simpos.root which contains generator level hit information. • CscClusterValidationOptions.py makes csc_clusters.root which contains detector simulated cluster information (So, correct DetDescrVersion should be crucial). • Then, run csc_cluster_performace.exe with two root files. It makes csc_perf.root which provides figures in the following pages. • Currently, job is run in lxplus.cern.ch. Their batch queue commands are same as BaBar’s (bsub, bjobs, klog,etc).

Cluster calibration: Qright/Qpeak vs. Qleft/Qpeak r-strips • Looks fine. CSS CSL phi-strips

• It’s saved in table and get a strip position by a quadratic interpolation. • Currently, it is used as it has in the code. • This can be generated by real data. • One could use his own table as below in jobOption.py file. # V02 MC QRAT calibration with nbin=50 and nrms= 2 CscThresholdClusterBuilder.qratmin_css_r = 0.0936681 CscThresholdClusterBuilder.qratcor_css_r = [ 0.000000, 0.000000, 0.000000, 0.000000, 0.0807402, 0.192581, 0.2788, 0.349086, 0.409295, 0.454935, 0.495891, 0.53068, 0.55803, 0.579625, 0.605338, 0.625196, 0.647682, 0.66527, 0.682893, 0.700273, 0.71683, 0.731681, 0.745759, 0.759357, 0.774508, 0.787322, 0.800335, 0.808468, 0.822776, 0.832892, 0.844826, 0.854548, 0.86595, 0.875246, 0.884159, 0.893254, 0.902383, 0.910208, 0.919269, 0.928248, 0.936756, 0.943576, 0.94925, 0.957871, 0.96772, 0.973168, 0.980292, 0.988297, 0.99183, 1.000000] # V02 MC QRAT calibration with nbin=50 and nrms= 2 CscThresholdClusterBuilder.qratmin_csl_r = 0.107754 CscThresholdClusterBuilder.qratcor_csl_r = [ 0.000000, 0.000000, 0.000000, 0.000000, 0.0398906, 0.0935101, 0.1996, 0.276374, 0.34136, 0.397006, 0.442904, 0.485013, 0.519259, 0.546959, 0.571139, 0.595666, 0.615712, 0.636289, 0.657206, 0.674757, 0.69259, 0.708839, 0.723793, 0.740804, 0.754768, 0.768269, 0.782545, 0.796264, 0.807948, 0.820126, 0.829775, 0.841212, 0.851852, 0.862724, 0.875594, 0.883446, 0.894059, 0.902656, 0.913039, 0.920321, 0.930196, 0.938684, 0.946356, 0.955538, 0.962261, 0.968984, 0.980072, 0.98636, 0.992682, 1.000000]

• Generator level hit information and cluster position in Clustering reconstruction are saved separately. Performace • Resolution is defined as the difference of two. The definition is a little different from residual. • The following figures are identical to Davids. – 10k single muminus in castor. Same MC to David’s. For muplus, it’s similar result as muminus. – In CscClusterization-00-10-20 rPull Sigma was:: 0.80 (vs 1.01) – Error of z position (dzc) may be updated in CscClusterization-00-10-21 ?? – No change observed in CscClusterization-00-10-22. RMS = 61 µ m σ = 47 µ m

Segment Performace (I) • The following figures are very close to David’s. – 10k single muminus in castor. Same MC to David’s. CscClusterization-00-10-22 is used rather than 00-10-21. • There are four parameters (r, r angle, phi, phi angle). – Either 2 X 2D segment or 4D segment algorithm available. r RMS = 31 µ m σ = 26 µ m Close to 47/sqrt(4)

Segment Performace (II) • The following figures are very close to David’s. – 10k single muminus in castor. Same MC to David’s. CscClusterization-00-10-22 is used rather than 00-10-21. • There are four parameters (r, r angle, phi, phi angle). – Either 2 X 2D segment or 4D segment algorithm available. φ RMS = 5.3 mm σ = 2.9 mm Different from 3.4/sqrt(4). Errors are correlated??

Segment Performance: Spoiled cluster multiplicity • There is correlation between real/fake muon and number of spoiled clusters. Real muon fake muon

In Real Data • The latest beam test data were taken from the CSC’s: – CERN (August 2004): x5 beam test – H8 (September 2004) – H8 (October 2004) • For x5 beam test data, it’s located at /castor/cern.ch/user/s/schernau/x5 with .csc format. • Castor :: CERN Advanced STORage System • These are readable by sitView software (MS Window based).

Resolution from Test Beam Data • They used three-point residual to estimate resolution. – The prediction is based on a straight line between layers 1 and 3, evaluated at layer 2. – x predicted = 0.5 (x 1 + x 3 ). – Therefore, the residual is given by r = x 2 – 0.5 (x 1 + x 3 ). – Error analysis relates the error in the residual to the error in the position measurement. – Assuming that each layer has the same resolution σ x , then the error in the residual is σ r = sqrt(3/2) σ x

• Different reconstruction method would result in different resolution.

• Higher counting rate will degrade CSC detector performance. • > 200Hz/cm 2 nominally expected in ATLAS.

• Cosmic ray data taking with 5 Toroid field on using Sector 13 4 6 3 7 in November 18-19 2006. 2 8 • Data will be taken for CSC detectors soon. 1 9 10 16 15 11 14 12 13

Possible Contribution • Data is usually saved in ByteStream format (binary file). Athena framework needs RDO format. BS should be converted to RDO when we analyze it in athena framework. – Unmanned and urgent (Vinnie’s comment). – To make a comparison b/w data and MC in athena framework, it’s essential. • Strengthen effort in CSC detector study. – Work together with David Adams (BNL). – Crucial code is already written in framework. – Currently, it’s validated only in single Muon MC sample. – Algorithm is much needed to be improved in battle environment such as charge correlation b/w x and y strips and timing information. – Israeli group (Tel Aviv Univ) are developing CSC algorithm for muonboy. They use HoughTransform algorithm.