workshop on b/tau physics, helsinki 30.5.-1.6.2002 v. karim ä ki, hip 1 software alignment of the...

Workshop on B/Tau Physics, Helsinki 30.5.-1.6.2002 V. Karimäki, HIP 1

Software Alignment of the CMS Software Alignment of the CMS Tracker Tracker

V. KarimV. Karimääki / HIPki / HIP

Workshop on B/Tau Physics at the LHCWorkshop on B/Tau Physics at the LHCHelsinki, May 30 - June 1, 2002Helsinki, May 30 - June 1, 2002

Topics:Topics: General considerationsGeneral considerations StrategiesStrategies AlgorithmsAlgorithms Helsinki auto-alignment algorithmHelsinki auto-alignment algorithm Concluding remarks Concluding remarks


General considerations


Alignment processes - hardware + softwareAlignment processes - hardware + software

Precisionassembly

Initial positions(modules)

Monitoring information

Positionscorrected

by monitoringAlignment by

tracks

Surveys in situ

For once

‘Final’calibratedpositions

Once a year …?

‘Continuous’?

For every shot?

Signal


Detector alignment by tracks - why, howDetector alignment by tracks - why, how

WhyWhy

Precision after assembly and optical surveys 0.1 - 0.2 mm (?)Precision after assembly and optical surveys 0.1 - 0.2 mm (?) Needs about 0.5 * hit resolution i.e. 5 - 30 Needs about 0.5 * hit resolution i.e. 5 - 30 mm Magnetic field effects, temperature effectsMagnetic field effects, temperature effects

HowHow Utilising natural smoothness of particle trajectoriesUtilising natural smoothness of particle trajectories

Using high pUsing high pT T trackstracks

Misalignments systematic offsets of hits from trajectoriesMisalignments systematic offsets of hits from trajectories

Software algorithm(s) to Software algorithm(s) to reduce systematic offsetsreduce systematic offsets by by

correcting the detector positionscorrecting the detector positions


Elementary alignment ‘manually’Elementary alignment ‘manually’

1D EXAMPLE


Importance of alignment precisionImportance of alignment precision

Example:Example:

Error in local u-coord due to tilt Error in local u-coord due to tilt

u ~ u tanu ~ u tan

u = 10 u = 10 mm = 45= 45oo

u = 5 cmu = 5 cm

Implies:Implies:

= 0.2 mrad = required angular precision= 0.2 mrad = required angular precision

u

Tilt angle


Algorithm developmentAlgorithm development

Formulations, basic tools:Formulations, basic tools: Dependence of correction parameters on residuals - can be Dependence of correction parameters on residuals - can be

complicatedcomplicated 2 2 minimization, Kalman Filter techniques, linear algebra - standard minimization, Kalman Filter techniques, linear algebra - standard

techniquestechniques misalignment tools (for development and validation)misalignment tools (for development and validation) track reconstructiontrack reconstruction

Algorithms validation:Algorithms validation: Monte Carlo simulation - comparison of known misalignments with Monte Carlo simulation - comparison of known misalignments with

corrections obtained by the algorithm - pull valuescorrections obtained by the algorithm - pull values Test-beam data - improvement of hit resolutions, trajectory qualityTest-beam data - improvement of hit resolutions, trajectory quality Alignment contest!Alignment contest!

referee referee to prepare recHit data set with misaligned detectorto prepare recHit data set with misaligned detector the algorithms should find the misalignments within errorsthe algorithms should find the misalignments within errors


CMS misalignment toolCMS misalignment toolHelge Voss

Britta SchweringTapio Lampen


Misalignment tool use casesMisalignment tool use cases

Study of misalignment effects on reconstructionStudy of misalignment effects on reconstruction Development and testing of alignment algorithmsDevelopment and testing of alignment algorithms

Movement rods/wedges x = y = z =1000 m:

Z events

Helge Voss:


Discussion of strategies


Hierarchy and application strategyHierarchy and application strategy

Hierarchy of alignment corrections:Hierarchy of alignment corrections:-full detector-full detector -barrel, F/B detectors-barrel, F/B detectors -barrel layers, forward disks-barrel layers, forward disks -barrel rods, forward wedges-barrel rods, forward wedges -detector modules (assumed planar)-detector modules (assumed planar)i.e. factorization of the problem when working on the alignment i.e. factorization of the problem when working on the alignment

Alignment correction mappings:Alignment correction mappings:-rotation, translation, 3+3=6 parameters per unit-rotation, translation, 3+3=6 parameters per unit-sag, twist, 2 or 3 parameters typically per unit (not for modules)-sag, twist, 2 or 3 parameters typically per unit (not for modules)

Application strategy:Application strategy:Compute and transform all global corrections down to the lowest Compute and transform all global corrections down to the lowest

level, i.e. to the detector moduleslevel, i.e. to the detector modules

Aligned reconstruction geometry = ideal geometry + wafers Aligned reconstruction geometry = ideal geometry + wafers position/orientation correctionsposition/orientation corrections


Software alignment strategiesSoftware alignment strategies

Distinguish:Distinguish:

1) Alignment start-up (launched at Day 1)1) Alignment start-up (launched at Day 1) input is ‘best geometry’ by assembly, survey and monitoringinput is ‘best geometry’ by assembly, survey and monitoring ‘‘large’ correctionslarge’ corrections steep (human) learning curvesteep (human) learning curve tedious, time takingtedious, time taking

2) Alignment calibrations (at regular intervals)2) Alignment calibrations (at regular intervals) input is previous best alignment + monitoring informationinput is previous best alignment + monitoring information repetition rate 1 day or more (experience only tells)repetition rate 1 day or more (experience only tells) ‘‘small’ correctionssmall’ corrections learning curve levelling off learning curve levelling off might become routine-likemight become routine-like yet always room for improvements (algorithms, statistics, …)yet always room for improvements (algorithms, statistics, …)


Possible schedule for alignment start-upPossible schedule for alignment start-up

Preparations:Preparations: Work out estimates of alignment errors (important for pattern Work out estimates of alignment errors (important for pattern

recognition)recognition) Further tuning of track reconstructionFurther tuning of track reconstruction Using isolated particles (high p_T muons)Using isolated particles (high p_T muons)Work inside out:Work inside out: Semi-independent alignment of Pixel detectorSemi-independent alignment of Pixel detector

tracks curvature by full tracker, especially for 2-layer Pixeltracks curvature by full tracker, especially for 2-layer Pixel vertex constraint important (therefore start with Pixel)vertex constraint important (therefore start with Pixel) hits only for Pixel independent alignmenthits only for Pixel independent alignment determines the coordinate systemdetermines the coordinate system given elements need to be fixed (simulations will help to tell us)given elements need to be fixed (simulations will help to tell us)

Continue with TIB, TOB, …Continue with TIB, TOB, … Matching between Tracker partsMatching between Tracker parts Matching with muon chambersMatching with muon chambers

Iteration cyclesIteration cycles


Sagitta and spurious sagittaSagitta and spurious sagitta

Sagitta: largest distance cord to arc

Spurious sagitta: change due to systematic errors in hit position =a measure of misalignments


Curvature accuracy for Pixel alignmentCurvature accuracy for Pixel alignment

SPURIOUS SAGITTA

In Pixel

In T

rack

er

• sagitta is proportional to L2

• consider 2-layer Pixel• sagitta error (Pixel) / sagitta error (Tracker) ~ (7.2/105)2 ~ 0.005

We conclude:

1 mm sagitta errordue to misaligmentsin full Tracker reflectsonly 5 m sagitta errorin 2-layer Pixel

We get precise enoughcurvature determinationfrom misaligned Trackerfor (semi)independentPixel alignment

Can we align Pixel independently of the rest of the Tracker on Day 1?Yes, if we can obtain precise enough curvature using full Tracker:


Candidate algorithms


Candidates for alignment algorithms - 1Candidates for alignment algorithms - 1

Alignment by Kalman Filter methodAlignment by Kalman Filter method Fruhwirth et al., CMS Note-2002/008Fruhwirth et al., CMS Note-2002/008 uses annealing to avoid secondary minimauses annealing to avoid secondary minima 3-parameter alignment tested with simulated test-3-parameter alignment tested with simulated test-

beam like setupbeam like setup

Helsinki auto-alignment methodHelsinki auto-alignment method 22 minimization formalism minimization formalism iterative, several passes over given dataiterative, several passes over given data up to 6 parameters per moduleup to 6 parameters per module successfully applied to real test-beam data (silicon successfully applied to real test-beam data (silicon

telescope) with 5 parameter alignmenttelescope) with 5 parameter alignment mathematics of the algorithm are simplemathematics of the algorithm are simple involves small matricesinvolves small matrices ORCA implementation under workORCA implementation under work


Candidates of alignment algorithms - 2Candidates of alignment algorithms - 2

H1 method (Gabathuler/PSI)H1 method (Gabathuler/PSI) used for H1 vertex detectorused for H1 vertex detector algorithm (implicit vertex constraint):algorithm (implicit vertex constraint):

all pairs of tracksall pairs of tracks minimizing sum of squared distances in space between minimizing sum of squared distances in space between

all track pairsall track pairs mathematics of the algorithm complicatedmathematics of the algorithm complicated

Blobel method (Raupach/Aachen)Blobel method (Raupach/Aachen) 22 solution of a very large number of parameters solution of a very large number of parameters CMS/Aachen people ...CMS/Aachen people ...

A large number of algorithms exist, more than A large number of algorithms exist, more than one per experiment, experiment specificone per experiment, experiment specific


Brief description of Helsinki algorithm


Helsinki auto-alignment - 1Helsinki auto-alignment - 1


Helsinki auto-alignment - 2Helsinki auto-alignment - 2


Concluding remarksConcluding remarks

Aiming at:Aiming at:Toolkit for tracker alignment with tracks = auto-alignment toolkitToolkit for tracker alignment with tracks = auto-alignment toolkitProviding an effective (default) algorithm Providing an effective (default) algorithm Stand-alone version for simulation of telescope-like setupStand-alone version for simulation of telescope-like setupTracker auto-alignment simulation in ORCA framework Tracker auto-alignment simulation in ORCA framework Status:Status:Mathematical formulation of an effective algorithm: derived, testedMathematical formulation of an effective algorithm: derived, testedUsed successfully in test-beam environment Used successfully in test-beam environment (CMS NOTE 2000/013)(CMS NOTE 2000/013)

OO modeling and coding of the 'core' classes: first iterationOO modeling and coding of the 'core' classes: first iterationSimplified telescope-like alignment simulation: first resultsSimplified telescope-like alignment simulation: first resultsInterface with ORCA: startedInterface with ORCA: startedFuture work:Future work:Development of ORCA interfaceDevelopment of ORCA interfaceStudies of auto-alignment for chosen parts of the TrackerStudies of auto-alignment for chosen parts of the TrackerStudies of using constraints (beam, vertex, $Z^0$-kinematics,\dots)Studies of using constraints (beam, vertex, $Z^0$-kinematics,\dots)Studies of correlated modules alignment Studies of correlated modules alignment etc...etc...

workshop on b/tau physics, helsinki 30.5.-1.6.2002 v. karim ä ki, hip 1 software alignment of the...

Documents