Minutes of the CALCOM analysis meeting, April 18, 1997. ====================================================== Attendance: (reconstructed after the meeting; sorry to those who attended but are not listed here) S. Barrow, W. Brooks, V. Burkert, A. Coleman, K. Egiyan, H. Funsten, J. Manak, B. Mecking, M. Mestayer, K. Mikhailov, B. Niczyporuk, F. Roudot, L.C. Smith, S. Stepanyan, S. Taylor, T.Y. Tung, D. Weygand. Agenda: K. Egiyan - Forward EC timing studies W. Brooks - Forward EC timing studies S. Stepanyan - Event reconstruction using Recsis and Simple Event Builder (SEB) B. Niczyporuk - Summary of the Hit Based Analysis and first results on DC time calibration F. Roudot - Comments on Recsis track reconstruction efficiencies -------------------------------------------------------------------------- Comments on Event reconstruction: ================================ (1) In Bogdan's analysis (see individual reports) a higher track reconstruction efficiency is achieved with more stringent cuts on the number of hits per superlayer (3 for SL1, 4 for SL2-SL6) than what Stepan gets using Recsis and SEB with looser cuts (2 for SL1, 4 for SL2-SL6). This discrepancy needs to be investigated further. Moreover, Bogdan's analysis takes 4 msec per trigger while Stepan reports 100 msec/event. Joe Manak will carry out a performance analysis of the two codes to understand this discrepancy. (2) Bogdan's results on the inclusive electron spectrum show the elastic/quasielastic H/C peaks (a CH2 target was used), while not showing any other structure. This differs significantly from Stepan's spectrum which, in addition to the elastic/quasi-elastic peak, also shows a pronounced peak at the Delta(1232) mass. The discepancy might be due to differences in the electron selection criteria (Stepan uses the Cherenkov counter information while Bogdan doesn't). (3) The pi-zero mass peak in Stepan's 2-gamma analysis is much wider than expected from simulations. Several effects may contribute to the discrepancy: (a) since only sector 6 was used in the analysis, the two gammas will probe mostly the outer periphery of the calorimeter where the EC stack calibration may not be well determined, e.g. attenuation lengths for short strips were not measured and are set in the data base to unrealistically high values. Also, the region close to the scintillator readout end may not be as well calibrated as the region further away from that end. (b) the expected mass resolution is a function of the pi-zero momentum while the simulations uses mono-energetic pions. Volker Burkert ----------------------------------------------------------------------------- Individual reports: K. Egiyan: ========== To obtain the time constants for EC strips the effective light path lenght (the effective light velocity) in the EC scintillators should be precisly defined. The ELV value was estimated from the hit path dependence of U-strip and the TOF scintillators time differences. The Time-Walk corrected data of the run 778 were used. A 15.5 cm/nsecvalue of ELV was obtained which is smaller then the expected 16-18 cm/nsec obtained in single scintillator strip measurements). The reason of the observed discrepancy will be studied. The study is hampered by the poor statistics of the data. W. Brooks: ========= A study of calorimeter timing resolution was presented. The data was from the February 1997 run. The method used was mean timing; this approach assumes a constant signal velocity for all stacks, but does not require that it be meas- ured. The main effort is to match phototube times for 216 phototubes, and to determine suitable time walk functions. The selection criteria for calibration events was that at least one tdc is reported for each calorimeter view in the inner and outer layers, and in a given view the strips hit are continguous (no gaps). (Single pixel events were not required.) The mean time in the inner or outer layer was defined to be the geometrically weighted average of the average times in each view; the average time in a given view was weighted by the number of photoelectrons. The time difference between the inner and outer mean times ('I-O') is independent of trigger time, and this quantity was used to match the 216 phototubes. The matching procedure was an iterative one, where I-O was cal- culated for each hit and a weighted correction factor was derived from it. The mean timing error can be estimated from I-O. All procedures were automated using PAW. The walk function and signal propagation velocity can be determined for each tube by playing inner against outer; i.e., the inner mean time can be used as a reference time to study individual outer tubes. In this approach, the calorimeter calibration does not rely on information from any other system. Results: this method applied to a MIP run (run 707) indicated the mean time error to be less than 900 ps, before any walk corrections are applied. The MIP run cannot be used to determine walk corrections because the pulse size range is too small. The above method applied to an 'electron' run (run 778) indicated the mean time error to be less than 1050 ps before walk corrections are applied. These numbers demonstrate that the iterative procedure works well for matching phototube times, and is insensitive to missing tubes (about 10 in this data). For this run the walk function was determined for each tube; the full range of the corrections was up to 10 ns as a function of ADC. Not all of the data was well fitted using the ADC variation; it is known that there is a significant walk correction associated with the time dispersion for light pulses travelling through long distances. A generalized walk function is needed which depends on both adc and distance from the phototube. This function will be derived from the existing data, however, more statistics will be needed to do an adequate job. The next step is to apply the measured walk functions to the data, which is expected to give a significant improvement; the ultimate timing is expected to be at the level of 200 ps for showering particles. B. Niczyporuk ============= a) The data, Run# 781.b01 (85072 triggers, Itorus = 1929A and Iminit = 6000A) has been processed with the SDA. The 13919 tracks (Q=-1) were reconstructed for Theta < 60 degrees. The 11235 electrons were identified by matching the DC tracks with the corresponding hits in SC and EC. The losses of 19% are mainly due to the geometry of EC. The corresponding MC perediction is about 14%. The resolution of track matching with electron cluster position is about 4cm for U-view and 5cm for V/W-views. The corresponding MC prediction is 3cm and 4cm. The uncorrected ratio R of energy cluster in EC to track momentum is: R = (Ecluster/Ptrack) = 0.2 +/- 0.034 The MC prediction gives: R = 0.218 +/- 0.019 The EC resolution is about twice worse (data: 17% and MC: 8.8%). As a first step one should more carefully examine the values of pedestals and gains used for this run. b) The hits in DC and SC of the identified electons were used to derive the calibration constants (time offsets:) for drift chambers, it is: Cdelay = DL1 - (Dcable)dc + (Dcable)sc [ns] where DL1 - Level1 trigger delay to stop DC tdc (common stop) (Dcable)dc - signal wire cable delay to start DC tdc (Dcable)sc - DC signal wire cable delay to stop SC tdc After corrections: for time of flight (ToF), time of propagation along SC slab, time of propagation along signal wire, and adding the raw tdc for SC hit with the raw tdc for DC hits on the reconstructed tracks (to get rid trigger jitter), the Cdelay has been determined with resolution of < 10 ns. Thus, having the DC filled with the slow gas (velocity about 20 microns/ns), the calibration constant resolution will contribute about 0.02 cm to the distance of closest approach (Dca). Having calibration constants (Cdelay) the drift times (Td) in a cell has been obtained. Next step, I am going to work on, is to derive: Dca = f(Td). Rest is ready to look the data with resolution very close to design value. One problem ----------- The drift time distribution Td in a cell of Region 1 DC exibits a dip at about 80 ns. This dip is also seen in raw tdc distribution. S. Stepanyan ============ Data from run 781 were analyzed using SEB (simple event builder) package in recsis after new demaxing (file clas_sector6_000781.b01). Analyses were done for two cases: 1)minimum required hits in the first segment of region 1 is 3 and it is 4 for other segments; 2)minimum required hits in the first segment of region 1 is 2 and it is 3 for other segments. In both cases selected electrons spectra were analyzed. In the W distribution on can see clear peak from elastic (quasi elastic) scattering of electrons. After the selection of events in the range of 0.35