Forum to exchange technical information to allow co-working on Monte Carlo Production for the test beam runs.

Updated from telephone meeting on 5-Feb-2008

present: Angela Lucaci-Timoce, Roman Poeschl, Felix Sefkow, Shaojun Lu, Frank Simon, David Ward, Nige Watson, Oliver Wendt

Responsible People

DESY: Oliver Wendt, Angela Lucaci-Timoce, adaptation of ILD scripts for job submission/book keeping
MPI: Frank Simon, Shaojun Lu, (gradually) take over and extend production system developed so far by Oliver et al.
Roman: Making digitisation/reconstruction job work on Mokka output
David/Nige: Overview/schedule, beam parameters, prioritisation of samples

Frank and Shaojun volunteer to take over running of MC production system as set up by Oliver, initially using the system that exists at DESY. In the future, the submission of grid jobs and running of MySQL databases to support them could be moved to another site (if there was some advantage) as the system consists of a small set of scripts/database tables.

Scripts/databases

Current status - summary slides from Oliver summary slides from Oliver, 5-Feb-2008 phone meeting

Oliver has adapted a cloned version of ILD's MC production system (Ivan Marchesini) for CALICE requirements, and agreed that scripts for submission continue to be run from within DESY. Can use either particlegun or stdhep file as input. Oliver is in process of further adapting the system to submit the combined digitisation/reconstruction Marlin job to the grid.

See also selected slides from Ivan Marchesini

Run Dependent Samples

Some parameters are specific to each run number (e.g. noise, gain corrections, ...), others apply to groups of similar runs (beam positions, energy spreads, Cerenkov pressure).

Run number specific parameters could be extracted from the Calice conditions database into steering/macro files for the Mokka generation and digitisation.

The run number to simulate (and the database tag immediately prior to start of MC production) could be written to the Mokka runHeader, and this could be unpacked by the digitisation processors to apply run-specific treatment of each sample.

See talks at CALICE software review for further discussion of this.

We agreed this is the ideal way to proceed, but may not be possible for very short term generation, e.g. for analyses aiming to produce results for CALOR'08. Shaojun would discuss with Oliver?

Separation of tasks

Agreed that running Mokka, digitisation of the Mokka output and reconstruction of digitised output should remain separate tasks (reco and digi separate Marlin processors), allowing revisions of each code separately. Expected that reco and digi would be run in a single Marlin job to run on the grid.

Official vs. user generated samples

The central MC production will be systematic, and details of code revisions used at each stage of processing, input steering files, logfiles and binaries used will be available.
Anyone may generate their own simulated samples of course, and is encouraged to make these public also. For compatibility, it would be sensible to start from the same steering files, etc., as for the official production runs. Proposals for improvements to the central production, e.g. better parametrisation of the beam profile, improved alignments, etc. would be very welcome.
We agreed that only "official", centrally produced MC samples should be used in analyses to be published, to ensure that Collaboration results are reliable and reproducible.
The CALICE Technical Board/Speakers' Bureau should consider whether this becomes official CALICE policy.

Input for Monte Carlo Production

• MC sample sizes per energy/position/particle type, 100-500k.

Suggestions for electron beam runs (DRW):

• runs.xls: Spreadsheet listing 2006 Electron runs at normal incidence
• Steering file commands to align the ECAL with the beamline detectors (August 2006)

        /Mokka/init/detectorModel TBCern0806_01_dchxy_new
        /Mokka/init/globalModelParameter EcalTranslateX  -27 mm
        /Mokka/init/globalModelParameter EcalTranslateY  -39.5 mm

• Steering file commands to align the ECAL with the beamline detectors (October 2006)

        /Mokka/init/detectorModel TBCern1006_01_dchxy_new
        /Mokka/init/globalModelParameter EcalTranslateX  -27.5 mm
        /Mokka/init/globalModelParameter EcalTranslateY  -41.1 mm

• Typical particle gun commands (with the numbers for position, position smearing and momentum smearing taken from my spreadsheet above).

       /run/particle/setCut 0.050 mm
       /gun/particle e-
       /gun/momentum 45 GeV
       /gun/momentumSmearing 0.098 GeV
       /gun/direction 0 0 1
       /gun/position  1.5 0.6 -60000 mm
       /gun/positionSmearing 11.0 4.8 0. mm
       /gun/info 

• I suggest tracking range cutoff of 50 microns (for the ECAL at least).
• See also MonteCarloSampleWishList - requests from analysers
  

Actions

• Roman - working on digitisation/reconstruction job for MC samples. To tell Frank and Shaojun where the current version of recon/digi job is so that they can start to familiarise themselves with CALICE s/w.
• Frank/Shaojun - contact their local grid guru and get themselves grid certificates, then to apply for CALICE VO membership, either via Roman or by http://grid.desy.de/certs/
• Oliver - continue adaptation of CALICE MC production scripts/database for submission of digi/reco jobs. Frank/Shaojun to discuss details of how running of system could be taken over from their user accounts, access control issues and so on (NKW: does DESY IT allow service accounts for such tasks, or is it always personal accounts only?).
  
• Query about whether updating a database from outside the site firewall is allowed?
• David - list of good runs for electron samples, to put onto web.
  

Earlier Samples Generated for 2006 data/LCWS'07

• See Lcws07Samples

User-Generated Samples

Various samples contributed by users, please add your samples here to share with others!

* Some files generated with a spatially uniform beam over the ECAL can be found at
/grid/calice/carlogan/tb2006/mc/FlatBeam

    the files are called G4_RunId_RandomSeed.slcio. 
    Random seeds 1000*** correspond to normally incident electrons, whereas RandomSeeds 3000*** correspond to electrons with a 7mrad tilt angle. 
    The energies follow the data energy for each run.
    The macro and steer files are in the subdirectory steer/
   

User-Requested MC Samples

People analysing the data are welcome to ask for additional samples to be generated, please add your requests to the "wish list"

MC mass production for test beam

• see MonteCarloFiles
• How to download the following MC mass production from GRID Storage elements (SE) --> HowtoDownload
• Please check out "/grid/calice/tb-MCProduction/" folder for the updated MC mass production.

Additional Geant4 (etc) modelling information

Some useful Geant4 modelling references in LCG-APP area, e.g. CERN-LCGAPP-2007-02, "Hadronic Shower Shape Studies in Geant4".