ties behnke: eu-lc simulation and reconstruction 1 eu-lc simulation & reconstruction full...
TRANSCRIPT
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
1
EU-LC Simulation & Reconstruction
Full simulation systems: status report
The next steps: where do we go from here
The simulation challenge: status report
The ECFA LC simulation group: Europe: David Ward, Ties Behnke :http://www-zeuthen.de/linear_collider
Ties Behnke, DESY
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
2
The mission
Coordinate the creation of appropriate simulation tools
Coordinate the creation of appropriate reconstruction tools
Provide a frame for the creation of appropriate analysis tools
The boundary conditions:
System should be long lived Should be lightweight We are not a collaboration: commercial software is difficult (funding) Share across regions as much as possible Little to no personpower available
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
3
The ingredients
GeneratorGenerator SimulatorSimulator ReconstructorReconstructor AnalyserAnalyser
Detector descriptionDetector description
“The data highway” IO (persistency) -- files“The data highway” IO (persistency) -- files
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
4
Full simulation
BRAHMS:GEANT3/ f77 basedTESLA TDR detector updated for recent changesComplete simulation (and reconstruction)
MOKKA:GEANT4/ C++ basedTESLA TDR detector “more or less”No reconstruction framework
BRAHMS is still the most complete package, but
MOKKA is quickly getting there, will soon replace BRAHMS SIM
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
5
The future of the simulator(s)
MOKKA is our main GEANT4 based full simulation environment
GEANT4 based light weight, C++ code geometry via mysql and C++ drivers
Our US collegues are evaluating MOKKA to use it as a base for their GEANT4 simulation environment.
MOKKA developments:
Improve the detector models allow for more flexible system: concept of user plugins Watch closely CERN geometry developments (GDML)
Should look at LCD-G4 – MOKKA and possible synergies / common developments
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
6
The data highway: LCIO
Need definition of a data model:
What is a hitWhat is a trackWhat is a calorimeter clusteretc.
Need a common way to store (“persist”) objects
JAS
ROOT
HBOOK
Objectivity
Anything else?
LCIO: a simple data model with underlying persistency system:
Serial packed files with pointer and “block” support
Event object
Hit objects
“Reconstructed Particles”: Track and cluster objects“Reconstructed quantities” objects
Com
plex
ity
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
7
LCIO: LC persistency framework
JETS
Vertex
HITS
Tracker hits
PARTICLES
Calorimeter Hits
Tracks
Clusters
OBJECTS
MONTE CARLO INFORMATION
SIMULATION RECONSTRUCTION ANALYSIS
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
8
LCIO data model
LCParticle
LCReconstructedParticle LCCompound
pointer pointer
specific extensions
type specificextensions
tracks
cluster
unique, complete
Kshortpion+
pion-
track 17
track 21
cluster 201Kshort2
The data model for reconstructed objects:
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
9
LCIO: Status
Developed in collaboration between SLAC / DESY / LLR (France)
Common data model / persistemcy framework for LC studies
Simple API to store and retrieve data
Same API in C++, Java (and Fortran)Simple underlying IO format (SIO): can be changed easily at a later time
Status:
Development started December 2002 First public pre-release in March First public production release for Montpellier: LCIO 1.0
Full C++ implementationFull Java implementationFull Fortran implementation Hit based data model “fully” defined, reconstructed objetcs still rudimentary
Status:
Development started December 2002 First public pre-release in March First public production release for Montpellier: LCIO 1.0
Full C++ implementationFull Java implementationFull Fortran implementation Hit based data model “fully” defined, reconstructed objetcs still rudimentary
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
10
LCIO Future developments
reconstruction data model
– refine data model as discussed …
– first implementation soon
add convenient methods and tools
– analyzing the MCParticle tree (graph)
● iterators for MCParticles, e.g. all stable/final particles
● analyzing parent/daughter relationships– adding convenience iterators for LCIO types
● simplify complex C++ syntax with dynamic_casts etc.
add possibility to store generic user data
– calibration constants etc.
respond to user requests
proposal by Frank Gaede
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
11
LCIO user extensions I
want to be able to store data not foreseen in LCIO data model– e.g. calibration constants, prototype data
impossible to extend the data model for all users’ needs !
existing LCIntVec and LCFloatVec not very convenient and efficient
still keep LCIO files self contained– i.e. data should be readable without additional software/ user code
user code independent of file format (SIO)
proposal by Frank Gaede
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
12
LCIO user extensions II
allow user classes with some restrictions:– subclasses of LCObject -> store in event collections– properties have to be basic types:
● int, float, double● can be extended to other types and vectors !
– Properties have to be accessible via user class’● data members or
● accessor member functions– Preprocessor macros for C++ and Java static functions for making class
known to LCIO
proposal by Frank Gaede
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
13
LCIO user extensions III
user classes are stored with data description in the LCIO files
data can be read back even if UserClass definition not available to the
program:
– LGenericObject:
– allows access to properties, via:
● names (convenient)
● indices (faster)
unknown collections can be analyzed with LCIO tools/browser
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
14
LCIO user classes, example I
class UserClass : public lcio::LCObject{ LCOBJECTINFO_FRIEND() ;protected: int myInt ; float myFloat ;public: UserClass(int i) : myInt(i) {} float getMyFloat() const; void setMyFloat(float f) ;} ;
#include "UserClass.h"
float UserClass::getMyFloat() const { return myFloat ;}void UserClass::setMyFloat(float f) { myFloat = f; }
// Make class known to LCIOLCOBJECTINFO_DEFINE( UserClass ) ;LCOBJECTINFO_PROPERTY_DIRECT( &UserClass::myInt , "myInt" ) ;LCOBJECTINFO_PROPERTY( &UserClass::getMyFloat, &UserClass::setMyFloat, "myFloat" ) ;LCOBJECTINFO_REGISTER( UserClass, "UserClass" ) ;
proposal by Frank Gaede
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
15
LCIO as online data format
Both TPC and calorimeter groups look at LCIO as online data format:
example: HCAL calorimeter group at DESY/Hamburg
calorimeter prototype: minical
data-MC comparison of energies in minical
First version of “raw data classes” have been defined.
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
16
Example: Calo hit file (online)
Run : 587 - MiniCal:Signature: HCRDHeadSize: 322Versions: 1|1|1|1RunNo: 587Person: sashaBeam: beamEnergy: 5Orientation: SIPM in frontEvents: 4000000Gate: 150StartUt: 1066153487Trigger: |0|5b|1|40|7|0|0|0|4c|fffffff3|ffffffff|ffffffbf|ffffffb8|ffffff8b|4|8|ffffffd0|4e|1|40|53|33|ffffff98|6|53|33|ffffff98|6|ffffffe0|fffffff3|ffffffff|ffffffbf|ffffff88|4d|1|40|ffffffd4|ffffff8f|4|8|ffffffe4|ffffffc8|1a|40|ffffffa0|ffffffa0|1a|40|c|27|9|40|ffffffb6|4|10|40|0|0|0|0|0|10|ffffffa0|0|3|0|0|0|22|0|0|0|ffffffff|ffffffff|ffffffff|ffffffff|0|0|0|0|ffffffb6|1|0|0|ffffffce|0|10|40|ffffffe4|ffffffc8|1a|40|ffffffa0|ffffffa0|1a|40|ffffffa0|ffffffa0|1a|40|NumADC: 9NumTile: 108X: 0Y: 0Top: 1500Bottom: 1500Veto: 0Comment:ADC_IDs: | 4| 5| 7| 8| 10| 11| 13| 14| 16|
Dump of data file:
detector : MiniCal
collection name : MiniCalRawADC
--------------- print out of CalorimeterHit collection ---------------
flag: 0x0 -> LCIO::CHBIT_LONG : 0 LCIO::CHBIT_BARREL : 0 LCIO::CHBIT_ID1 : 0 LCIO::CHBIT_PDG : 0
cellId0(bytes) | cellId1(bytes) | energy/amplitude | position (x,y,z)0: 1/1/15/1 | 15/0/0/0 | -0.0037418 | ( no position avaliable ) |1: 1/2/15/1 | 15/1/0/0 | -0.00508885 | ( no position avaliable ) |2: 1/3/15/1 | 15/2/0/0 | -0.0105953 | ( no position avaliable ) |3: 2/1/15/1 | 15/3/0/0 | -0.0164467 | ( no position avaliable ) |4: 2/2/15/1 | 15/4/0/0 | -0.0418246 | ( no position avaliable ) |5: 2/3/15/1 | 15/5/0/0 | -0.0189564 | ( no position avaliable ) |6: 3/1/15/1 | 15/6/0/0 | -0.0130033 | ( no position avaliable ) |7: 3/2/15/1 | 15/7/0/0 | -0.0447505 | ( no position avaliable ) |8: 3/3/15/1 | 15/8/0/0 | -0.000621088 | ( no position avaliable )
LCIO can be used online
eases access to data and comparison to MC
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
17
Reconstruction
BRAHMS/ f77 based reconstruction:
Complete and sophisticated trackingParticle flow reconstruction packageTools (n-tuple for analysis, interface to main packages like ZVTOP, ...)
OO world:
No coherent reconstruction frame yetLCIO allows using f77 based reconstruction package on MOKKA output
(first version of MOKKA- BRAHMS connection released)
We need to move fast towards a OO based reconstruction framework, to enable reconstruction program developments!
We need to move fast towards a OO based reconstruction framework, to enable reconstruction program developments!
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
18
Reconstruction
BRAHMS: sophisticated and complete reconstruction software availableOO based: no reconstruction tools available at the moment
Status of reconstruction packages:
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
19
Reconstruction
Hits simulated in detector (BRAHMS)
Tracks and Pflowobjects reconstructed(BRAHMS reco)
Plots Karsten Buesser
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
20
The Plan
Develop a plain C++ reconstruction framework
LCIO as data modelPossibly CGA as geometry access modelVery simple interface, no fancy interactive environment for a startMulti language support is no problem (JAVA ...)
Use LCIO ROOT LCIO JAS
to provide a simple user interaction with the reconstruction
At this stage: remain independent of any particular environment (root, JAS, PAW, ...)be open for future developmentsbe light-weight (minimise the number of libraries the user needs to install)
At this stage: remain independent of any particular environment (root, JAS, PAW, ...)be open for future developmentsbe light-weight (minimise the number of libraries the user needs to install)
.... but we are still far from this goal...
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
21
Fast Simulation
SIMDET: fast, parametrised Monte Carlo
NEW: interface to used track based ZVTOP in SIMDET improvements to parametrisations
SGV: fast, semi-parametrised Monte Carlo (produces hits)
NEW: interface to ZVTOP in SGV
At the moment, fast simulations work, are producing physics resultsbutin Europe we have no project for a modern fast simulator
Our US collegues are developing a similar system
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
22
Software: Outlook
Simulation:
Converge towards GEANT4 based MOKKA simulation program should become more flexible and user friendly Close collaboration with our US collegues on simulation establised
IO:
LC specific persistency scheme and data model established Interface to C++, JAVA and Fortran exists Common development with our US collegues
Reconstruction:
Next big project Digitisation and reconstruction “complete” in BRAHMS Much effort needed in OO framework: the next project
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
23
Towards a simulation challenge
Software environment: see previous slides
Access to data:
Need transparent and flexible access system Data volume small compared to LHC, but still significant
The plan:
User interface/ interaction: use the GRID computing model and software
Storage manager: dCache (FNAL and DESY, integrated into GRID)
GLOBUS authentication manager
The challenge: be able at the end of the current ECFA study to do a “blind” analysis challenge (A. Wagner, Amsterdam)
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
24
LC Data depository
Database interface to data stored under dCache at DESY
Current problems: access from outside DESY (being worked on)
Tie
s B
ehn
ke:
EU
-LC
S
imul
atio
n a
nd R
econ
stru
ctio
n
25
Summary
Software tools present a significant challenge
Progress on the simulation: GEANT3 based simulation more stable and completeGEANT4 based simulation becoming availableCommon persistency scheme is reality
Progress on the reconstruction:BRAHMS based reconstruction existsOO reconstruction still in the futureLCIO input existsLCIO output being defined
Goals: GEANT4 based simultorSimple reconstruction environment independent of particular frameworkLCIO as basic data format and model to facilitate data exchangeManagement of data sets through GRID like system