Transcript MEG Software Status

MEG Software Status
Framework for MC,
Schedule,
DC reconstruction update
and
Discussion on a ROOT-based offline
MEG Software Group
MEG-Review Feb. 2004
1
Question from reviewer
• Give a common goal to be achieved by the
next review;
“I have appreciated the presentation of the offline
software team, with a team leader etc... As
promised I come forth with the question: could you
give us -- and yourself -- a common goal to be
achieved with that software in time for the next
review?”
MEG-Review Feb. 2004
2
Our goal was:
• MC unified framework
– With double option: Tokyo or Pisa
• Large Prototype MC unification
– CEX beam test
– Same offline package for DATA & MC
• Event Display
– Initially GEANT3
MEG-Review Feb. 2004
3
MC Framework for MEG
• Unified framework with double option:
–
–
–
–
–
Tokyo or Pisa code.
Tree structure & skeleton,
Merged materials & tracking medium,
Stand alone & built-in event generator,
DC, TC, Magnet and B field are merged.
• Framework based on:
–
–
–
–
Fortran 77,
CERNLIB,
ZEBRA I/O,
GEANT3.
• Distributed via CVS.
MEG-Review Feb. 2004
4
To be done: MEG MC
• I/O
– ZEBRA structure … in progress
– Digitization
• Liq. Xe
– Geometry … in progress
– Scintillation photon tracking … in progress
• Full reconstruction program
• Offline database… ODB? Separate
database (e.g.PostgreSQL, MySQL, etc.)?
MEG-Review Feb. 2004
5
Large Proto CEX beam test
• MC merge into the new framework
… just finished.
– Common output format for DATA/MC
– Same offline package for DATA & MC
– Scintillation photon tracking in Liq. Xe
… Full ray tracing or analytical tracking
– Rayleigh scattering/absorption in Liq. Xe
– Reflection on the PMT quartz window
… Fresnel or total reflection
– Scintillation light spectrum
… Monochromatic, Gaussian, Basov et.al, etc.
MEG-Review Feb. 2004
6
To be done: beam test
•
•
•
•
NaI response simulation
LiH target and Vessel simulation
Phase space simulation for p0
Offline database … ODB? Separate database
(e.g. PostgreSQL, MySQL, etc.)?
MEG-Review Feb. 2004
7
Schedule/man power
• By end of April 2004
– Liq. Xe geometry merging … S.Yamada/F.Cei
– ZEBRA data structure … P.Cattaneo
• By end of June 2004
– MC output with Digitized waveform …
XE: S.Yamada/F.Cei, DC: H.Nishiguchi, TC: P.Cattaneo
• Summer DC beam test will performed using
unified MC
MEG-Review Feb. 2004
8
DC reconstruction
• Track Dictionary
– Efficiency
– Resolution
MEG-Review Feb. 2004
9
The dictionary concept
Several tracks with
similar kinematics
producing a single hit pattern
Hit pattern
Single and unique string
(i.e. a dictionary key)
Average (over the set)
MEG-Review
Feb. 2004
track
parameters
10
Building the dictionary
• MC sample used to build the dictionary:
– Positrons from Michel decay;
– Unpolarized muons;
– Generator level cuts:
0.08 < |cosθ| < 0.35;
-60° < φ < 60° .
• No Tdrift used,
• No Z measurement used yet.
MEG-Review Feb. 2004
11
250000 generated events 12900
patterns; efficiency = 95%
The population of
uniform:
40%
43%
13%
4%
the patterns is not
has 1 entry
has 2 ÷ 10 entries
has 11 ÷ 50 entries
more than 50 entries
Number of events in a dictionary record
MEG-Review Feb. 2004
12
Momentum components in the dictionary
LEFT
RIGHT
Event by
event
distributions
Px / MeV
Average
in each
Dictionary
record
Py / MeV
Py / MeV
A Hit pattern
cannot tell the
sign of PZ
Pz / MeV
Pz / MeV
Px / MeV
Track first turn
has hits in at
least three
sectors
MEG-Review Feb. 2004
13
Dictionary “resolution”
Generate a sample of
independent events
vertex X
vertex Y
Px
Py
For tracks in the dictionary
acceptance (Nsectors > 2)
find the dictionary key
compare;
Px with <Px>(key);
normalize to RMS<Px>
pMC - < pdict >
σ
14
vertex Z
Pz
To be done: dictionary
• Optimize stats. given by 1 – eff. ~ 10-3 - 10-4 and
by looking at RMS vs. stats. (intrinsic resolution
of method);
• Add noise hits;
• Add inefficiency of Drift Chamber;
• Add drift time;
• Superimpose tracks.
MEG-Review Feb. 2004
15
Discussion on a ROOT-based
offline system
• Fully OO framework, including:
– all needed functionality present (from data taking to final
plots, geometry package with integrated event display)
– transparent LAN, WAN and HPSS support
– parallel computation (PROOF)
– interface to SQL/RDBMS
– interface to GEANT3 with migration tools
• Extensive CERN support
• ROOT I/O exceeds MEG requirement (raw
data throughput: ~3.5MB/s)
– Max ROOT I/O throughput: ~51MB/s(from FNAL test)
• Scalable to cope with trigger rate uncertainties
MEG-Review Feb. 2004
16
MONARC for computing & data model
• A central site, Tier-0
– Prompt calibration, reconstruction and run info
• Regional centers, Tier-1/2
– Reprocessing
– MC production
– Mirror locally the reconstructed objects
• Tier-3/4 centers
– Analysis
• Offline system reads raw data streams via
rootd and tree output streams:
– Raw event database (to HPSS)
– Tag database (on disk)
– Run Catalog database (on RDBMS)
17
The Run Manager and the VMC
Detectors
DCH
TOF
EMC
STEER
run management
interface classes
detector base classes
data structure base classes
EVGEN
PYTHIA6
HIJING
MEG offline
main program
…
VMC
Geant4 VMC
Geant3 VMC
ROOT
Geant4
Geant3
MICROCERN
External
packages
18