MEG Software Group Shuei YAMADA 3 Feb 2006
MEG Software Group Shuei YAMADA 3 Feb. 2006 1
Contents Summary of Software Meeting (3 -4 Nov. 2005) Standard Operating System Transition to Subversion ROME based analysis tools Status of Software Organization Status of Monte Carlo Status of Offline Software Ongoing Studies Using MC Resources and Needs CPU DISK Network Schedule and Manpower Shuei YAMADA 3 Feb. 2006 2
Summary of Software Meeting (3 -4 Nov. 2005) Standard Operating System Scientific Linux (SL) Mostly used in the HEP community Derived from Red. Hat Enterprise Linux, Freely available Transition to Subversion A concurrent versioning system, similar to, but better than CVS Truly atomic commit (CVS: file-by-file) Moving/renaming files and directories (CVS: loses history) Free/Open source version control system Runs on all modern flavor of Unix, Mac & Win 2 k/XP Binary package available for SL 3/4 Shuei YAMADA 3 Feb. 2006 3
ROME based analysis tool Meg. Root was rejected Analysis tools based on ROME was approved megbartender : event cocktail & digitization meganalyzer : reconstruction & event display Both for online & offline New software coordination Repository MC Offline Online Shuei YAMADA : Fabrizio Cei, Shuei Yamada : Matthias Scheebeli, Ryu Sawada : Stefan Ritt 3 Feb. 2006 4
Status of Software Organization Monte Carlo Status Offline Software Status Ongoing Studies Using MC Shuei YAMADA 3 Feb. 2006 5
Software Organization Simulation Analysis Shuei YAMADA DC H. Nishiguchi TC P. Cattaneo XE F. Cei + S. Yamada Beam/Target W. Ootani + K. Ozone + V. Tumakov Calibration F. Cei DC digitization P. Huwe TC digitization P. Huwe XE digitization Y. Uchiyama Trigger simulation D. Nicolo’ + Y. Hisamatsu Framework M. Schneebeli Database R. Sawada DC H. Nishiguchi + M. Schneebeli TC D. Zanello XE G. Signorelli + R. Sawada 3 Feb. 2006 6
Procedure of Analysis experiment simulation DAQ GEM MC simulation ZEBRA MIDAS bartender ROOT analyzer ROOT Shuei YAMADA event generation detector simulation Electronics simulation event cocktail waveform simulation digitization trigger simulation Reconstruction Event display 3 Feb. 2006 7
Monte Carlo Status GEANT 3 based simulation program : GEM Program built around the framework REM Organization in modules, as an OO class; structured like: xxx (prefix of specific device) = dch, ticp, ticz, xec … yyy (suffix defining the functionality) = ini, set, end, draw, … Documentation under SVN in meg/rem/doc Interactive GEM Call GEANT 3 functions interactively to Draw geometry, track and hit Change the running conditions To be done: Learn how to use it Implement user interface (kumac and/or GUI) Shuei YAMADA 3 Feb. 2006 8
Drift Chamber Progress Improved integration of time to distance profile from GARFIELD GEANT simulation of Vernier electrode patterns Improved hit structure More adequate description w/ low-E -rays Simple and small hit data structure (220 → 8 words/hit) Efficient (80 → 100%) and simple hit cell# calculation To be done Effect on signal of Vernier patterns Improve detail of material : electronic cards, cables, … Shuei YAMADA 3 Feb. 2006 9
Timing Counters Progress: Implementation of many geometries : cables, jacket, bars w/ slanted ends, square fibers, PMTs, APDs, … Preliminary support structure Generation and propagation of scintillation photons : based on analytical calculation & Poisson fluctuations To be done: Improve details of materials and support structure Improve photon propagation model Cross check w/ beam test data (for counter) Shuei YAMADA 3 Feb. 2006 10
Liquid Xenon Calorimeter Progress: Geometry almost finalized Inner&Outer vessel, PMT holders, PMT position, Honeycomb, … Faster GEANT based optical photon tracking : (~10 s/event on 3 GHz Pentium 4, ~ 7. 5 s/event on 2. 4 GHz Athlon 64) To be done: Implement final geometry : Dense PMT layout for backside, hollow spacer for L&R side Fast optical photon tracking Shuei YAMADA 3 Feb. 2006 11
Beam and Target Progress: Muon beam simulation based on phase space into event generator Elliptical tube option for target Implemented end cap and Rohacell insertion tube To be done: Implement Beam Transport Solenoid Complete the beam transport within the detector Implement final target with support 3 D Field map + interpolation Shuei YAMADA 3 Feb. 2006 12
Calibration Extensive work in the past Partially implemented in MC Progress: Geometry and tracking media almost completed Event generation under testing To be done: Test and commit event generator Complete implementation of geometry Ni plates outside LXe calorimeter (neutron calibration) Shuei YAMADA 3 Feb. 2006 13
MC General Conclusion Simulation status satisfactory : good level of sophistication in geometry and physical process simulation. Further refinements in geometry under way. Simulation of calibration procedures started. Some people (5 - 6) actively working. It’s a good time for testing the mass production (possible problems with disk space, memory management…) Shuei YAMADA 3 Feb. 2006 14
Offline Software Status Single Event Display Preliminary mu -> e+gamma trigger simulation Waveform simulation & digitization XE & DC : ready TC : preliminary Analysis Extensive works by individuals Partially implemented in meganalyzer : XE : Fast reconstruction (Qsum, position, …), Position reconstruction DC : preliminary track fit (for online display) Shuei YAMADA 3 Feb. 2006 15
Single Event Display ROME and ARGUS are merged Both for online & offline ROME runs in 3 modes ROME stand alone Argus stand alone ROME with Argus display (new mode) Configuration in the ROME configuration file No special user code needed Shuei YAMADA 3 Feb. 2006 16
QT movie… Shuei YAMADA 3 Feb. 2006 17
Ongoing Studies Using MC BG Source Study Optimize end cap , target system and Rohacell tube design AIF events in LXe Source of AIF gammas’ spectrum & yield AIF Rejection LXe Waveform simulation Pile-up rejection Shuei YAMADA 3 Feb. 2006 18
BG Source Study New features Purpose Optimize end-caps Upstream End-cap DC cable duct Rohacell Insertion Tube Michel decay beam Optimize target system Shuei YAMADA 3 Feb. 2006 19
BG from End Cap 39 photons / 50, 000 Michel e+ SUS beam Michel decay Designed end-cap (Aluminum part) + Bremsstrahlung photon Aluminum SUS Beam Pipe Shuei YAMADA 3 Feb. 2006 20
BG from Rohacell Tube Rohacell X 0 = 820 cm density = 0. 052 g/cm 3 C 9 H 13 O 2 N Bremsstrahlung photon + 3 photo Shuei YAMADA ns / e Michel 0 0 0 , 50 3 Feb. 2006 21
BG from DC cable duct Carbon fiber Aluminum Cu cable Bremsstrahlung photon beam Michel decay 427 photons / 50, 000 Michel e+ Shuei YAMADA 3 Feb. 2006 22
BG Source Study(1) Summary BG Sources Effect of cable ducts was improved down to 1/2 Bremsstrahlung & Annihilation at rest Effect of end-cap is relatively small Low E (~ 1 Me. V), but pile-up study is needed AIF study target & Rohacell tube Cable duct Target study Energy deposit in LXe [Me. V] Slant angle of the target Complete target support structure & calibration system Shuei YAMADA 3 Feb. 2006 23
AIF study using MC 1. Generate Michel e+ in target, emit them for 4π 2. At the each GEANT tracking step, calculate annihilation probability by material information and Michel e+ momentum information 3. Generate 2 γ at each step weighted by this probability 4. Trace 2γ, if it enters Xe cryostat Shuei YAMADA 3 Feb. 2006 24
AIF probability map No Cut X-Y View Z-X View Xe events Egam> 45 Me. V Shuei YAMADA 3 Feb. 2006 25
AIF Spectrum & Photon Yield AIF spectrum and their origin Egam Me. V/52. 8 Me. V Shuei YAMADA Photon yield per muon decay Egam Me. V/52. 8 Me. V 3 Feb. 2006 26
AIF Event Rejection in LXe Different arrival times of 2 gammas Different Impinging points δT 2 gamma [sec] TODO : Pattern recognition Shuei YAMADA δX 2 gamma [cm] 3 Feb. 2006 27
LXe Waveform Simulation Pile-up rejection Take sum of PMT outputs Larger pulse Microstructure in pulse shape disappear Optimization for # of PMTs to be summed Summing up all the PMTs not good from S/N viewpoint Shuei YAMADA 3 Feb. 2006 28
Xe Waveform Simulation 1. Sum up Gaussians over all photo electrons • • Mean = arrival time of each photon Width = TTS (1 p. e. response) Fits very well with real pulse ! 2. Shaping (Low Pass Filter) • Low Pass filter • Time constant RC = 5 nsec Shuei YAMADA 3 Feb. 2006 29
Pulse Shape Fluctuation Pulse shapes are not constant especially for small pulses data simulation 0. 6 V 1. 2 V deviation Fluctuation reproduced Shuei YAMADA 3 Feb. 2006 30
Pile-up finding Peak search DT=50 ns. DT=15 ns Count # of peaks in Moving average Simple but powerful for large DT Differential Count # of peaks in Differentiation Powerful for small DT 11 Me. V + 42 Me. V Shuei YAMADA 3 Feb. 2006 31
Rejection Efficiency V 2 )/5 Optimal Value : 60 ~ 70% of Qsum Function of Energy Shuei YAMADA (E 1 3 Feb. 2006 + E 2 e. 8 M 32
Rejection Efficiency As functions of Energy of each g and DT Summed up to DT 8 ns 60% Qsum Misidentification probability: <0. 05% DT 50 ns Weak points: DT < 10 nsec Small pulse after a large one DT 10 ns DT 100 ns DT 15 ns Shuei YAMADA 3 Feb. 2006 33
Resources & Needs Data storage CPU Power Data Access Shuei YAMADA 3 Feb. 2006 34
Data Storage Resource & Needs PSI Tapes total PSI Disks MEG Needs 30 -40 TBytes (free) + 40 TBytes occupied by back up (to be freed) 4 TBytes (backuped) ~ 10 TBytes/yr (read data) + 6 TBytes(not backuped) ~ 40 -50 TBytes/yr (MC Production) ~ 10 TB/yr (overheads, DSTs) 70 -80 TBytes 10 TBytes 60 -70 TBytes/yr MEG Needs ~ 5 TBytes/year of Disk Space for DATA Assuming 1/2 of the data collected in one year reside on disk for monitoring, calibrations, faster analysis, etc… Shuei YAMADA 3 Feb. 2006 35
CPU Resource & Needs PSI Nodes MEG Needs (CPUs/yr) 64 ~ 3 CPUs (real data, w/o Waveform fitting) < 1 CPU (selected data w/ Waveform fitting) ~ 20 CPUs (MC production & bartender) ~ 10 CPUs (MC reconstruction = 3 x data, w/o WF fitting) < 1 CPU (MC selected sample w/ WF fitting) 128 CPUs Total Shuei YAMADA 64 CPUs ~ 33 (+20 per 10 repr. ) CPU/yr 3 Feb. 2006 36
Data Access Resource & Needs PSI Link Speed 25 MBytes/s to tapes via FTP 1 Gbits/s to disks from CPUs MEG Needs ~ 1 MBytes/s (w/ Waveform compression) ~ 10 MBytes/s (w/o Waveform compression) OK ! Shuei YAMADA 3 Feb. 2006 37
Schedules and Manpower Milestones Manpower Shuei YAMADA 3 Feb. 2006 38
Milestones Within 2 -3 weeks New PSI cluster partly ready : 128 Opterons MC mass production (at least signal events and Michel positrons) Start development of reconstruction & Pattern recognition algorithms Start pre-selection study By end of September Finish MC mass production Signal, Michel positrons, backgrounds… Shuei YAMADA 3 Feb. 2006 39
Schedule for LXe analysis Other sub-detectors can emulate LXe schedule Shuei YAMADA 3 Feb. 2006 40
Man Power 2006 Q 1 Y. Hisamatsu 0. 5 H. Nishiguchi 0. 2 W. Ootani 0. 1 K. Ozone 0. 5 R. Sawada 0. 5 Y. Uchiyama 0. 7 S. Ritt 0. 1 M. Schneebeli 0. 4 F. Cei 1. 0 G. Gallucci 1. 0 D. Nicolo’ 0. 2 A. Papa 0. 3 R. Pazzi 1. 0 G. Signorelli 0. 5 P. W. Cattaneo 0. 5 D. Zanello 0. 5 A. Barchiesi 1. 0 W. Molzon 0. 2 V. Tumakov 1. 0 S. Yamada 1. 0 P. Huwe 1. 0 F. Xiao 0. 7 Total 12. 9 Shuei YAMADA 2006 Q 2 Y. Hisamatsu 0. 5 H. Nishiguchi 0. 3 W. Ootani 0. 1 K. Ozone 0. 5 R. Sawada 0. 5 Y. Uchiyama 0. 7 M. Schneebeli 0. 8 F. Cei 1. 0 G. Gallucci 1. 0 D. Nicolo’ 0. 2 A. Papa 0. 3 R. Pazzi 1. 0 G. Signorelli 0. 5 P. W. Cattaneo 0. 5 D. Zanello 0. 5 A. Barchiesi 1. 0 W. Molzon 0. 2 V. Tumakov 1. 0 S. Yamada 1. 0 P. Huwe 1. 0 J. Perry 0. 7 D. Stute 0. 4 F. Xiao 0. 5 Total 14. 2 3 Feb. 2006 41
End of Slides Shuei YAMADA 3 Feb. 2006 42
- Slides: 42