Minutes of the TEST BEAM ANALYSIS meeting of 1 June 1999 (Prague) 0) Summary of May 99 testbeam data Clara Troncon May 99 testbeam was very succesful: useful data have been taken with - 3 full size modules for the first time with both frontends: FE-B and FE-C, flex and bare - 17 single chip assemblies with both frontends, FE-B and FE-C various sensor designs: SSG,SNB,ST1,ST2,SAN sensors prototype 1 and 1b, 300 and 200 micron thickness, some with MCMD processing variouos bump bonders (IZM, ALENIA, BOEING) 3 irradiated sensors at different running conditions (voltage, shaping time, threshold, detector angle). Results from data analysis of 7 single chip assemblies and 1 module were presented at the meeting. I) Status of May 99 testbeam analysis Christoph Linder reported about the status of the MCMD analysis in Wuppertal. They succeeded to use the program h8 to align and produce ntuples and plan to investigate the behaviour of routed pixels for the three different routing schemes implemented in columns 2 and 3 (connection to neighbour pixel, to the second neighbour (skip 1), to third neighbour (skip 2)). Single hits population seems to show an excess in columns 2-3, but this has to be investigated systematically. The residual of routed pixels shows a displacement from 0, as expected (proper decoding has to be provided). X-Position inside pixels for double hits clusters has similar distribution for either routed or not routed pixels. Stephan Meuser reported about the status of FEC analysis in Bonn. They succeeded to use the program h8 to align and produce ntuples. - A global resolution of 13 um was shown for a 300 micron thick SSG1b sensor, while the 2 pixel cluster resolution was 7.4 but showed a second peak at -50 um. - The TOT distribution was shown for single (gaussian distr.) and double (Landau distr.) pixel clusters. - TOT profile inside pixels was flat in x, showed losses at the edges in y (even and odd rows folded). - Attilio Andreazza studied the distance between pixel hit position and extrapolation both for FEB and FEC and found that for FEC the distribution shows some peaks on the negative side at -100, -200, -400. The hit probabilities vs extrapolation position inside pixel show anomalies when extrapolation misses the hit by -1 or -3 pixels (+1,+2 don't show anomalies). FEB does not present this behaviour. - Interesting plots from test beam monitoring program on module M3 (FEC, bare, SSG 1) Francesco Ragusa reported about the status of analysis in Milano. - FEC efficiency: 1) constant during the burst 2) 3 % of hits in next BX (Beam crossing) 3) 2 pixels cluster residual shows a 2nd peak at -50 um, diplaced by 1 row. The second peak is removed by requiring the same lvl1 for the 2 hits. Same effect seen last year in FEA. Maybe related to large TOT. 4) Problems with large TOT--> row address --> lvl1 ? An understanding of a possible read out error is needed to study timing performaces. - ALENIA/FEB efficiency 97.1 % Looks reasonable, given that the sensor was an old ST2 design and measurements were done at high threshold. - Charge collection of sensors prototype 1b. Measurements done with 4 assemblies: 1) SSG 1b FEB (200 um) CIS-IZM calibrated, slow shaping, low threshold 2300 e- 2) SNB 1b FEB (200 um) CIS-IZM calibrated, slow shaping, low threshold 2300 e- 3) SSG 1b FEC (200 um) CIS-IZM not calibr, 1.6 us/ threshold=3000 e- 4) SSG 1b FEC (300 um) CIS-IZM not calibr, 1.6 us/ threshold=3000 e- some saturation seen on single hit clusters in FEC Prototype 1b behaves as expected, SSG: only ~ 10 % charge loss extending +-30 um around the grid, SNB : no loss. - Efficiency Module M2 (FEB,flex, sensor prototype 1b, 200 micron thickness) This analysis required a major effort, to extend to a full module the software developed for single chip assemblies. Good efficiency: 98.4 %, flat plateau. 0.7 % of 0 hits (masked pixels not corrected, chip 5, accumulation near pixel edges (threshold)) 0.7 % not on time (low PH, near pixel edges) Mario Aleppo (Milano) presented a new measurement of the Lorentz angle for the IT2 sensor irradiated at 0.5e15 n/cm2 at 600 V bias voltage. He uses the mean cluster multiplicity as a function of angle. The Lorentz angle is given by the minimum of this distribution. Two fitting functions were used: a simple parabola and a function extracted by a Monte Carlo simulation developed for this purpose. It has as input the parameters we measured or used in the test beam (diffusion, depletion depth,threshold). The two results agree very well and give a Lorentz angle of -3.1+-0.4+-0.2 degrees. To be compared with a value of -3.1+-0.4+-0.6 degrees he obtained last year for IT21 (1.e15 n/cm2) and a value of -9.1+-0.4+-0.5 degrees for a not irradiated detector with 98 test beam data. Torsten Harenberg (Wuppertal) presented a new idea to be used to measure the Lorentz angle. It is based on the measurement of cluster center shift from the expected position due to the magnetic field. It needs a very accurate track fit. Results on data are expected soon. II) Resolution Telescope resolution: George Gollin (CPPM) presented his studies to optimize the pointing accuracy of the silicon strip telescope used in the H8 test beam, as a function of the telescope geometry and material. He wrote a Monte Carlo simulation and found out (one of) the best extrapolation accuracy on Pixel plane is obtained when a)the two pixel planes are placed respectively one between the two first strip planes and the other between the last two strip planes b)pairs of strip detectors as close to each pixel layer as possible. This solution has the advantage to provide reasonable tracking accuracy even when the second or third strip plane is malfunctioning and was adopted during the May 1999 testbeam, taking care to minimize the amount of material between the first and last strip plane. The predicted extrapolation error is 3.4 (3.8) microns, for a 150 Gev/c pion beam (predicts 4.9 for september 98). Pixel resolution: Clara Troncon compared the measurements of pixel intrinsic resolution obtained in 1998 and 1999 testbeam. The two hit clusters position was evaluated both using a binary algorithm (the measured coordinate is taken at the center of the two hit pixels) and an analog algorithm (the position is interpolated using the charge collected by the two pixel using the Time Over Threshold measurements). The best result (4.4 um) is obtained for all the three 200 um thick sensors tested, for the 2 hit cluster residual. This value is dominated by/could be considered as a measurement of the telescope extrapolation error and was never obtained before (last year best value was 5.5 um) then confirming the real improvement of the telescope tracking performance in 1999. The 300 um thick sensor gave a corresponding value of 4.9 um, for the 2 hit cluster residual. The difference between the values obtained for the two thicknesses is explained in terms of a reduced charge sharing region (we measured +-4 um vs +-5-7 um) in 200 um sensors, and this is due to an interplay between threshold and reduced total charge and reduced thickness (less diffusion). Since the fraction of two hit cluster is also reduced (10 % vs 16 %) the global resolution is dominated by 1 hit clusters and is 13 um at 0 degrees. A larger optimal angle in 200 um sensors than in 300 um sensors can also be anticipated. The resolution of the highly (e15 n/cm2) irradiated IT21 sensor (1998 data) was also presented as a function of incident angle, using both analog and digital algorithms. A value not larger than 15 um was found between 0 and 30 degrees. III) Status of the analysis tools and organization Analysis program: Francesco Ragusa summarised the most relevant new features of the alignment and reconstruction program: new Sirocco format, new calibration file format, Module decoding, track quality diagnostics and possibility to use 3 hit tracks. The program can be found in: /afs/cern.ch/atlas/maxidisk/d43/h8tb/milano/programs/h8-08.06.99.tar.gz To be done: by: 1) New Routine for hot strip George Gollin + ... 2) New geometry routine for module/orientation ... 3) Ganged pixel clusters + multichip clusters Mario Aleppo + GeGo 4) MCMD geometry decoding & reconstruction Christoph Linder 5) Extend classes of tracks for alignment ... 6) Include multiple scattering GeGo 7) Test/debug the new version with the single chip assemblies FrRa + MaAl + ... Calibration files, data base of configurations, masks: The laboratories involved, i.e. LBL,Bonn,Genova,Wuppertal will provide calibration of the assemblies in the same conditions (voltage, shaping time, temperature if irradiated) used in test beam, as well as masks. A repository will be created (by C.T.) on /afs/cern.ch/atlas/maxidisk/d43/h8tb/calibrations where all these files will be kept. Raw data of May 98 test beam available at: Marseille, CERN tapes, Milano. Ntuples available on request. Organization: No central production of ntuples, at present, since still in debug. Central repository of: recontruction program, calibration files and masks, database of configuration files. Code developed by users Compatibility (new names for new variables, INCLUDE ? in kumac/functions) IV) Future test beams : programs, needs, organization Pierre Delpierre summarised the most relevant new features of the testbeam setup: new mechanics with easy rotation survey, new DAQ hardware,i.e faster SIROCCO allowing to collect 600 events/burst, revised software for module DAQ, improved monitoring (Module +event display), new arrangement of strips and pixels planes. Then schedule and program were discussed: next test beam from July 21 till 4 August. Several (~ 15) single chip assemblies and (~8) modules expected.