muon detector description status and plans
DESCRIPTION
Muon Detector Description status and plans. summary of contributions from Laurent Chevalier, Andrea Dell’Acqua, Jean Francois Laporte, Stefania Spagnolo, Dennis Wright. Outline. Geometry tags from Atlas-CSC-01-02-00 (Bulk of CSC production) to Atlas-Geo-00-00-00 (data taking layout) - PowerPoint PPT PresentationTRANSCRIPT
12/06/2008 Muon Week - CERN 1
Muon Detector DescriptionMuon Detector Descriptionstatus and plansstatus and plans
summary of contributions from
Laurent Chevalier, Andrea Dell’Acqua, Jean Francois Laporte, Stefania Spagnolo, Dennis Wright
12/06/2008 Muon Week - CERN 2
OutlineOutline
– Geometry tags• from Atlas-CSC-01-02-00 (Bulk of CSC production) to Atlas-Geo-00-00-00 (data taking layout)
– XML to GeoModel translation for Inert Materials– Performance issues:
• Robustness: clash cleanup • Memory allocation
– FDR2: alignment constants loaded run-time• Atlas-CSC-02-00-00 and Atlas-CSC-02-01-00 the FDR 2 tags
– Summary of Muon DD functionality in recent and future releases
12/06/2008 Muon Week - CERN 3
Content/use of a Geometry tagContent/use of a Geometry tag
* Structured according to amdb specific conventions
** Derived(not primary) from XML
amdb blob = dump of the ascii file(active + XML dead matter) in a blob
XML
(*)
amdb blob
tables(*)
tables(**)
asci
i file
Muonboy via Amdc
Id Dict. name
Cabling Maps names
DigitizationData DecodingReconstruction
hopefully soon
12/06/2008 Muon Week - CERN 4
Geometry tagsGeometry tags• ATLAS-CSC-01-02-00 ATLAS-CSC-01-02-00
– Based on Technical Drawings of Layout R (amdb version R01) but • Station positions (P-lines) account for Egg shape global deformation
– default station position ≠ nominal layout
• Additional random displacements (~1mm) and tilts (~1mrad) of stations with MDTs as static Alignment Corrections (A-lines)
– A-lines loaded init time from the same source of P-lines (geom-DB, no Cond-DB) but applied as corrections (Delta Transforms) on top of the default position and orientation of the station
– MuonGeoModel implementation fully driven by primary numbers in the Oracle tables
• (=primary source) for the active part (good validation)• (≠primary source) for the inert materials (very limited validation)
– Barrel Toroid, Endcap Toroids, Atlas Feet, Shielding (disk, forward and toroid)
– many missing blocks Calorimeter Saddle, TGC supports, Big wheel supports … …others, some not yet implemented in Amdb
ATLAS-CSC-01-00-00
12/06/2008 Muon Week - CERN 5
Geometry tagsGeometry tags• ATLAS-GEO-00-00-00 ATLAS-GEO-00-00-00 vs ATLAS-CSC-01-00-00
– Moving from Layout R in tech. drawings to the MS layout as installed (amdb version R03)
– Active Geometry:• MDT HV-RO side checked and fixed for a few chambers• station positions fixed for a few stations• RPC layout in Small Sectors + RPC positioning in the stations with single RPC
units + RPC chambers with cutouts + RPC special chambers heavily revisited • CSC position fixed (layer 2 installed not layer 1)
– Dead Matter:• In the XML description: fixes to MDT Big Wheel, Calorimeter Saddle, TGC
support; pos. of ID Patch Panels PP2; Muon Barrel + sect. 13) Access Platforms, Feet Struts, pos., BOL Support, Tie road, ECT cryostat fix
– Validation of R03: VERY LIMITED UNTIL NOW – NEED HIGH STAT. SAMPLEs • hit relocation OK – Amdc vs MGM OK – No unexpected clashes involving active
geo. • first simulation tests showed a problem with CSC now fixed • for reconstruction/trigger (LVL1 + HLT) emulation we need new RPC cabling
maps (few chambers - BOG and S2/3 in BMS at 2 and 4- with #of strips changed w.r.t. old layouts)
12/06/2008 Muon Week - CERN 6
RPC layout in R03RPC layout in R03 As shown also from hit residuals
studies with cosmic ray data, the RPC layout in the small sectors does not correspond to the nominal layout
ATLAS-CSC-01-00-00
ATLAS-GEO-00-00-00
see B. Liberti et al http://indico.cern.ch/getFile.py/access?contribId=0&resId=0&materialId=slides&confId=32282
12/06/2008 Muon Week - CERN 7
RPC layout in R03RPC layout in R03
BMS5BMS2A/CX
BMS4BMS3A/CX
BMS6BMS4A/CX
s2 RPCchambers
s3 RPCchambers
changed w.r.t. R01 in •position in the station•# of eta strips •cutouts
BOG as an example of chamber with single RPC unit:in some cases inverted position of gas gaps and thick support panel
12/06/2008 Muon Week - CERN 8
Dead Matter in GeoModel/GEANT Dead Matter in GeoModel/GEANT from XMLfrom XML
Andrea Dell’Acqua developed a XML -> GeoModel parser for a transparent translation of Amdb dead matter in GeoModel – no more painful hand coding of the dead matter shape and positioning– use directly primary source of dead matter description
AGDD2Geo in 14.2.0 (not active yet in any default configuration)• Amdb blob (with XML dead matter) read from Oracle• all components in Amdb are translated into GeoModel volumes,
hence Geant4 volumes via Geo2G4• by job-option one can enable to built a selection; no changes needed
for the moment in MuonGeoModel • planning to start with MDT/TGC supports in the big wheels
– if tests in simulation are OK add platforms, services at eta 0, calo saddle, etc.
– presumably wait longer before replacing components already existing in MuonGeoModel
12/06/2008 Muon Week - CERN 9
AGDD2GeoAGDD2Geo
MDT supports
AGDD2Geo required - some fixing of syntax of XML description in amdb (done in r03)- an update in Geo2G4 in order to handle assemblies of assemblies (done for 14.2.10)
12/06/2008 Muon Week - CERN 10
AGDD2GeoAGDD2Geo
TGC supports
12/06/2008 Muon Week - CERN 11
AGDD2GeoAGDD2Geo Access platforms
Active geo. (MuonGeoModel) + new dead together
12/06/2008 Muon Week - CERN 12
AGDD2GeoAGDD2GeoTO DO in 14.2.0 and 14.2.10• material consistency checks: AGDD had extra materials with nocorrespondence in Oracle yet. • geometry checks (mostly in simulation)check for unforeseen overlaps, for correct material distribution• GEANT4 tracking performance maybe the most critical chapter (impact on simulation timing, memory)• reconstruction checks make sure the whole chain from simulation to reconstruction works properly
After these checks – we’ll freeze, as default, a configuration of default options for MuonGeoModel + AGDD2Geo to enable the tested components in simulation and reconstruction
12/06/2008 Muon Week - CERN 13
Clash review – Clash review – simulation robustnesssimulation robustness • Dennis Wright working on a very careful clash hunting and fixing• fixes in 14.2.0
– started from nominally aligned layouts, cured residual small active-active clashes due mostly to shared surfaces (introduced suitable tolerances on volume dimensions)
– moved to misaligned layout: many clashes of tilted chamber corners (~100 m) cured by reshaping a tighter station envelop, by subtracting and adding half cylinder shapes to a minimal trapezoid
• event display becomes heavy • solved by enabling(in MC)/disabling(in Reco) this fine clash fixing procedure by
job option
– simplified spacer beam description + fixes (should improve G4 tracking perf.) + refine LB and CR description linking one object to the other to (close to reality)
– fix BOG cutouts – attaching active-inert clashes
• fixed feet description to fix a clash with BOG chambers
work in r03
12/06/2008 Muon Week - CERN 14
Clash review – Clash review – simulation robustnesssimulation robustness • after 14.2.0 working on cutouts:
– currently only cutouts in MDTs of BOG stations are implemented
– strategy to implement the rest:
• use GeoCutVolAction for RPCs and DEDs(class propagates cutouts in mother volume down to all daughters)
– OK for above components because there aren't many sub-volumes
• must handle MDTs differently to avoid creating too many boolean volumes
– re-writing Mdt.cxx and parts of MultiLayer.cxx in order to generalize MDT cutouts to other cutout chambers besides BOGs
• improvements in AMDB required in order to continue
work in r03
12/06/2008 Muon Week - CERN 15
Memory issues Memory issues • MuonGeoModel budget ~ 130 MB (out of ~1GB
allocated by simulation tasks, ~2GB allocated by reconstruction tasks)– 50 MB are purely from GeoModel
• can be reduced to a minimum in reconstruction by forgetting all the internal details of detectors (i.e, components linked to readout geometry classes, used by reconstruction)
– to be done
– 80 MB are a tracking related cache (used to be built run-time on demand, since 14.1.0 is fully built init time)
• can be turned on in simulation– easy and done
• can be optimised (centrally, not in MuonGeoModel)
12/06/2008 Muon Week - CERN 16
FDR2: FDR2: alignment constants loaded run-alignment constants loaded run-timetime
MuonAlignmentDbTool
loadParameters
Cool FolderMUONALIGN/MDT/BARREL
Cool FolderMUONALIGN/MDT/BARREL
MuonDetectorTool::Align
registered for call back against
MUONALIGN/MDT/BARREL
A-line container
in storegate
A-line container
in storegate
MuonDetectorManagerupdateAlignment
compute deltaTransformsclear tracking cache
update GeoModel transform tree
clear/update MuonGeoModel tracking cache
step 1
step 2
The Alignment wg provided Alines in COOL computed with an emulation of the
optical al. system with IoV in the range of the FDR2
data sets just in time for the FDR2 start-up
System tested up to reconstruction ~1 year ago with old
COOL versions and run/event based IoV
definition
Before FDR2 we have been able to test the access to the
condDB by emulating fake events with time-stamps
matching the A-lines IoV --- test OKbut no integration of
reconstruction
unable to test with coherent sim. data + cond. data in advance
12/06/2008 Muon Week - CERN 17
FDR2: FDR2: alignment constants loaded run-alignment constants loaded run-timetime• The integration of reconstruction has been tried just at FDR2 –
we found out what we missed– by default MuonGeoModel assumes no access to the cond DB
• minimal change in configuration• tested on run 522090, before trying in FDR2
– worked fine if MuonTrackingGeometry disabled (otherwise a susp. crash)
• integrated: but crash run-time on a data set 522080 disable – Of course testing in final full environment are mandatory to ensure
no bugs are lurking around• a bug identified in MuonGeoModel in the mechanism for clearing the
tracking related cache – maybe related to the observed crash – MuonAlignmentDbTool/MuonDetectorTool are designed to read a
single folder from COOL when updated• but Optical alignment for Barrel / EndCap A / EndCap C are independent
systems writing asynchronously into COOL !• some redesign is needed
– Better planning + better definition of responsibilities would have helped: goals of FDR2 for the muon alignment wg and subsequent requirements on muon software discussed in ~April but not updated as needed up to FDR2 start
12/06/2008 Muon Week - CERN 18
FDR2: FDR2: alignment constants loaded run-alignment constants loaded run-timetime
– From the test on run 522090, before trying in FDR2• with MuonTrackingGeometry disabled and the extra conf. options Full reconstruction has been run and the effect of loading/not-loading the
alignment constant was observed in muon reconstruction
d0condData-d0def.reco /mmbarrel
d0condData-d0def.reco /mmendcaps
MuonBoy reconstructed
tracks (MS hits only) back extrapolated to IP
rms = 470 mNo diff. in the
output of reco. in ECs since only
barrel A-lines were read from
COOL and passed to
Muonboy by theMuonGeoModel MuonDetectorM
gr
No diff. in the output of reco.
in ECs since only barrel A-lines
were read from COOL and passed to
Muonboy by theMuonGeoModel MuonDetectorM
gr
12/06/2008 Muon Week - CERN 19
Summary of Muon DD functionality in Summary of Muon DD functionality in recent and future releasesrecent and future releasesIn 14.2.0 / 14.2.10
– MuonGeoModel can handle old layouts + ATLAS-GEO-00-00-00 (much closer to ATLAS pit for active geo. than before) still need extensive validation
• make sure we handle it in reconstruction (cablings updates for RPC !)
– small differences in conf. for simulation and reconstruction (default): • caching flag OFF, fine clash removal ON
– default dead matter as before (a part from minor fixes)– new dead matter (via XMLGeoModel parsing) will be validated in custom
productions
In 14.1.0.Y (FDR2 reconstruction release) and AtlasPoint1 branch– Try to finalize the access and use of condition data (access to more
folders, full test up to reconstruction) for FDR 2 samples and for cosmic data (A-lines measured in P4/5 – available for barrel sector 5 and EC)
By 14.3.0 integrate validated new dead matter and alignment data handling
12/06/2008 Muon Week - CERN 20
ConclusionsConclusions• Lot of progress in many areas:
– new dead matter– systematic clash fix
• Data force to improve realism:– RPC layout, CSC layout
• some revision of TGC probably necessary
• Data and full scale exercise (as FDR2) demonstrate that progress/dev. of sw components is driven by muon wide objectives
improve communication
from a recent example: from FDR2 planning to Alignment wg and software wg (geometry + db access + checks are reco. level)
• Still needed – better way of enforcing coherence: simul / reco / cablings svc.s – better definition of intermediate task “borders” better
scheduling of work to be done