1. Report on the Beam Test Analysis Satoru Uozumi Apr-25 2007 GLDCAL meeting Topics of this talk are: Comparison of various MIP calibration methods Some problems …

2. The DESY Beam-line ECAL module Movable stage e + 1~6 GeV Veto1 Drift chambers DESY II Electron Synchrotron e-e- Drift chambers Trigger-1 Trigger-2 & Veto-1 ECAL Veto-2 e + beam T1 T2 Veto2 T3

3. Cuts on analog signal of the Trigger and Veto counters are introduced to select certain 1 MIP events. MIP Event Selection (Trigger & Veto) Reject Not used

4. MIP Event Selection (cont’d) Black … trigger/veto cuts Red … Yellow strips have non-pedestal signal Blue … Green strips have no signal

5. Extracting MIP calibration constants Four methods examined : 1.Fit around the MIP peak with asymmetric Gaussian (  is used as the calibration constant). 2. Fit around the MIP peak with Landau function convoluted with Gaussian (Most Probable Value (a) is used as the calib. const.). 3. Just take a mean value of the selected MIP events. 4. No calibration, just use uniform calibration constants, (but do perform inter-module correction).

6. Method 1 : Fit with Asymmetric Gaussian fit region :  -2.5  1 ~  +2.5  2 Fiber modulesDirect modulesKNU modules Typical fit Worst fit Typical fit Worst fit Typical fit Worst fit

7. Method 1 : Fit with Asymmetric Gaussian (cont’d) Calibration constants With 1 st configuration

8. Method 1 : Fit with Asymmetric Gaussian (Inter-module correction) In the longitudinal shower curve, there is a gap at the border of different module types. To correct this effect, “Inter-module correction factor” f inter-module is applied to calculate the total measured energy: f inter-module is calculated to have the best  /E in each beam energy. 1. 6 GeV 1 st config 1 GeV Inter-module correction Gap

9. Method 2 : Fit with Landau x Gaussian fit region : (0.5 ~ 2.5)xMPV (fiber module), (0~3)xMPV (other two) Fiber modulesDirect modules Worst fit Typical fit KNU modules

10. Method 2 : Fit with Landau x Gaussian (cont’d) Calibration constants With 1 st configuration

11. Method 3 : Just Take a Mean Value An apparent problem : Remaining pedestal events will affect to the mean value. Calibration Constants with 1 st configuration

12. Method 4 : No calibration (but perform inter-module correction)

13. Energy Resolution with 4 different calibration constants (1 st config, energy-dependent inter-module correction)

14. 1 GeV2 GeV3 GeV4 GeV5 GeV6 GeV No calib. 14.15+0.0910.02+0.058.23+0.037.29+0.036.60+0.026.09+0.02 Asymm. G 14.03+0.0910.04+0.058.34+0.037.48+0.036.83+0.026.34+0.02 Landau x G 14.00+0.0910.00+0.058.30+0.037.45+0.036.81+0.026.33+0.02 Mean 13.92+0.099.92+0.058.20+0.037.34+0.036.69+0.026.21+0.02  stat  const No calib.13.81 + 0.072.30 + 0.10 Asymm. G13.53 + 0.073.12 + 0.07 Landau x G13.47 + 0.073.13 + 0.07 Mean13.43 + 0.072.92 + 0.08 Unit of  /E and  is per cent. Table of  /E Fit results with Energy Resolution with 4 different alibration constants (1st config., energy-dependent inter-module correction) Problem 1 : Why no calibration case shows the smallest  /E in E >= 4 GeV ? Problem 2 : Why constant term is so large in all the cases ? - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

15. Linearity with 4 different calibration constants (1st config, energy-dependent inter-module correction) Problem 3 : What is the cause of this non-linearity? Energy-dependent (or over-estimate of) inter-module correction ? Also it could be due to the temperature variation ? Of course the MPPC saturation effect should be considered, too.

16. Problem 4 : Bumps on Longitudinal Shower Shape No calibration (but with inter-module correction) c.c. from Meanc.c. from Asymm. G. Some bumps observed, even after the calibration. Bump on layer-6 looks like over-correction, but others are not… Currently no idea how to solve. 1 st configuration 1 GeV center injection

17. Appendix : A sign of strip response non-uniformity (1 st configuration, c.c. from mean value) Injecting position : X and Y layers show different level of outputs due to non-uniformity of the strip response.

18. Summary Several types of MIP calibration methods are examined, but none of them seems to be perfect. There are several problems found : –Effort on calibration do not improve  /E –  constant is not small (2~3 %) –Non-linearity –Bumps on the longitudinal shower curve I personally suspect all of these problems are linked to almost one origin, the MIP calibration and the inter-module correction. More study is necessary, but what we should do for the preliminary result toward coming workshops ?