Calibration and alignment software Marian Ivanov Outlook n
Calibration and alignment software Marian Ivanov
Outlook n n n Impact of systematic effects on physical results TPC calibration TPC alignment
Statistical uncertainty n n n R-Phi and Phi resolution for perfectly aligned and calibrated TPC (at the TPC entrance) Given by the cluster position resolution ( divided by sqrt(Npoints)) At low momentum – influence of the multiple scattering
Misalignment of detectors n Linear misalignment can be detected by our algorithm Statistic of 2000 tracks per sector (IROC+OROC) ( 72000 tracks) is big enough to be on the level below statistical uncertainty n Tested with stand-alone (fast) simulator n n Following slides – precision of the alignment parameter determination for two different statistic sets
Track fitting n Ali. Rieman used for track fitting n n Picture: n n Less than 1 s for track fitting (20000 tracks) Pt resolution for non aligned sectors Input misalignment n n 2 mm in translation 1 mrad rotation 1/ptrec-1/pt
Results –Rotation Z n n Left side – 2000 track samples Right side – 5000 track samples
Translation X n n Left side – 2000 track samples Right side – 5000 track samples
Translation Y n n Left side – 2000 track samples Right side – 5000 track samples
Result (Pt residuals) n Relative pt resolution (dpt/pt) n n Left side before alignment Right side after alignment
Alignment - Ex. B n Ex. B effect – simulated – linear dependence expected n n Xshift = kx*(z-250) – kx=0. 005 Yshift = ky*(z-250) - ky=0. 005 The same in both sectors Alignment with tracks (2000 track samples) n Systematic shifts in translation estimates (negligible in comparison with statistical error) n n X – 0. 02 mm, Y – 0. 08 mm, Z – 0. 003 mm Systematic shift in rotation estimates n Rz – 0. 05 mrad, Ry – 0. 006 mrad, Rx – 0. 006 mrad
Warning example - STAR - TPC Grid. Leak distortion n n Dependence on field, track charge, location, luminosity consistent with ion leakage at gated grid gap Hopefully not the case of Alice TPC
Alice Ex. B distortion (M. Kowalski) n Radial distortions at lower and outer TPC radius due to the nonuinformity of magnetic field – E field perfectly aligned with B field at central membrane n n Alice - Omega tau – 0. 354 (E=400 V/cm, B=0. 5 T) Note : n n Non linear as function of z Phi dependence
Alice Ex. B distortion (M. Kowalski) n Azimuthal distortions at lower and outer TPC radius due to the nonuinformity of magnetic field n n Dy = 90 cm x 0. 0018 ~0. 16 cm (STAR reported magnitude of correction on the level ~0. 1 cm – nucl-ex/0301015) Systematic error - 4 times bigger than statististical
Alice Ex. B distortion n Influence n Systematic effect to the DCA resolution n The influence on the pt resolution will be estimated n n n The distortion z and theta dependent For the first analysis the cut on the DCA has to be adjusted Realistic magnetic field description needed (see next slides) Track finding efficiency in TPC should be not be affected – (Ex. B distortion is a smooth function) Influence on the TPC-ITS track matching
L 3 field components Tesla calculation (M. Losasso) currently in Aliroot I = 30 k. A
L 3 field components Measured field, I = 30 k. A (from ntuples of A. Morsch) No corrections for possible probes misalignment applied
Drift velocity q Requirements (systematic error on the level of statistical error) q q q Measurements q q n Z resolution ~ 0. 01 cm vdrift precession ~ 0. 4*10^-4 Drift monitor – GOOFY ~ 10^-4 Tracks crossing central membrane STAR TPC n q (Initial) drift velocities determined / monitored with lasers Automated updating of drift velocities (and initial T 0) from laser runs q q Checked / fine-tuned by matching primary vertex Z position using east and west half tracks separately (Alice – algorithm tested by C. Cheskov) Ideally determined from track-matching to SVT (perpendicular drift), but requires all other calibs to be done already! (principle has been tested)
Electron attachment n n n Electrons can be absorbed in the gas during the drift The probability to be captured by an O 2 molecule is 1% per 1 m drift per 1 ppm of O 2 (NA 49) Alice – expected oxygen content (ALICE MC)~ 5 ppm n n Should be achieved (Joachim) Influence n n n Non systematic effect to the position resolution Affects only statistical uncertainty by a factor sqrt(absorbtion) and d. Edx measurement Does not affect multiplicity measurement
Gain calibration n The chip gains vary in range of 5% n n n The gain variation due to electrostatics (for example anode wire sagita) n n Expected cluster position variation on the level of 0. 05* pad width Expected random behavior does not affect the cluster position – (the effect of local variation of gain is negligible as compared to cluster size) Influence: n n Small influence on the pt resolution and efficiency d. Edx affected
TPC calibration: Outlook n n n TPC calibration parameters TPC calibration classes MI approach: n n The size of the calibration data in CDB (Condition Database) and in memory (during reconstruction) dominated by the size of data for pad by pad. Everything else negligible. Store all data which can be used in the reconstruction, respectively which can used to indicate problems. n n Particularly the data from the sensors (voltages, currents, temperature sensors) Offline code status
Calibration classes n Ali. TPCCal. Det n n n Calibration parameters specific to each sector: One array of 72 floats Ali. TPCCal. Pad n Parameters specific to single Pad: n n Gain. Factor, T 0, Pad Response Function Width, Noise Used to pattern local variations of detector parameters One array of 72 Ali. TPCCal. ROC objects Ali. TPCCal. ROC n n Actual container of single ROC specific data One array of [Nchannels] floats Nchannels depends on the type of sector in stack (inner, outer) Interface n n Ali. TPCCal. ROC(Int_t sector) Set. Value(padrow, pad, value) Get. Value(padrow, pad) Memory consumption n Npads x sizeof(value ) n n n 0. 5 million channels * sizeof(value) 1 D array for each sector Mapping index – (padrow- row) using external map array class Ali. TPCRoc (1 per outer sector, 1 per inner sector)
TPC calibration parameters –per pad Parameter N. of Unit channels Source Update frequency Gain factor 557568 Relative Offline/HLT Rare Time 0 557568 Relative ? Offline/HLT Rare Preamp-shaper width 557568 Relative ? Offline/HLT Rare Noise 557568 Relative (sigma) ? Rare n The difference between relative and absolute is in the data volume n n n ~ 2 MBy relative ~ 8 MBy absolute Current implementation in Ali. Root – use floats
TPC conditions – per set of sensors Parameter Temperature probes N. of Information Source Update frequency channels ~4500 sensors on Array of : ID, FEC, snesors on position, samples space frame? (temparature) frame? ? ? temparature) in Interface to DCS time ? Array of : ID, samples (voltage and DCS current) in time Per run Drift voltage (VHV) ? Array of : ID, samples (voltage and DCS current) in time Per run Gating voltages ? Array of : ID, voltage n Per run ? Per surveyer measurement Avoid problems with versioning Define queries Data volume depends on the sampling frequency n n Array of : ID, position, angles DCS The format should be defined as soon as possible n n Per run High voltage Laser parameters n DCS and ? Can be reduced by fitting The data format and functionality – Not TPC specific n n Common class should be defined Request for offline group presented (Hopefully someone will implement it)
TPC calibration parameters – per TPC Parameter N. of Information Source Update frequency channels Oxygen content 1 Samples in time DCS Per run Drift velocity monitor (Goofy) 2 Samples in time DCS? Per run
Altro setup Parameter Data Source Update frequency volume Altro frequncy 0 Altro acquisition window 0 Moving average (on/off) 0 Zerro suppresion (on/off) 0 Tail cancelation (on/off) 0
TPC calibration parameters – per TPC Parameter n n Data Source Update frequency volume Drift velocity map (parameterization) ? Offfline Rare Space charge map ? Offline Rare Ex. B correction map ? Offline Per change of magnetic field The above result in the distortion map The data volume depends on the grid size
TPC parameters for reconstruction Parameter Data Source Update frequency volume Signal shape parameterization (diffusion parameter) 0 Offfline Rare Local error parameterization () 0 Offline Rare
Shuttle Schema n Ali. Shuttle – The Shuttle n Ali. Shuttle. Config – Interface to n Ali. DCSClient – Provides DCS n Ali. Shuttle. Trigger – Interface to program manager. Organizes conditions data retrieval, preprocessing and storing it to CDB. the configuration stored into LDAP server API. Communicates with DCS AMANDA server over TCP/IP DAQ Log. Book and client to DAQ “End of Run” notification service
Offline calibration - Status n Calibration classes for pad parameters implemented Default parameters stored in the database n Pad gain variation (+- 5%) n n n Used in simulation and reconstruction Noise, T 0, and Preamp shaper width - will be implemented soon in the simulation n Typical variation of parameters needed as input
Alignment - Outlook n n n Toy model results presented in previous slides Short overview of reconstruction framework (Cvetan Cheskov) Current development n Implement alignment algorithms inside of Ali. Root alignment framework
Alignment framework n n Space-points extraction and processing (filtering) Track fitting Track extrapolation points Residuals minimization
Framework Overview 1/2 Reconstruction Phase I ESD file with track space-points Distributed Phase II Tree with Selected Space points Local file Phase III Phase IV Build tree index Alignment procedures Local
Space-points retrieval (Phase I) n During the reconstruction, in between backward propagation and refitting: n Loop over ESD tracks and sub-detectors (ITS, TPC, TRD, TOF, RICH): n n n Get cluster indexes Call trackers to get the space points Store the points inside the ESD track n The storage of space-points is controlled by Ali. Reconstruction: : Set. Write. Alignment. Data() n Unified Ali. ESDtrack method of getting #clusters and their indexes: n n Get. Ncls(Int_t i. Det) & Get. Clusters(Int_t i. Det, UInt_t*) Abstract method of Ali. Tracker: n n Get. Track. Point(Int_t index, Ali. Track. Point &p) Method implemented for ITS, TPC, TRD, TOF
Space points filtering (Phase II) n Filtering: Take the ESD trees in a TChain n Select on ESD track parameters n Store selected space point arrays into tree (in local file) for further analysis n n n So far a simple (local analysis case) ESD processing is implemented A TSelector prototype is being implemented (distributed analysis case)
Framework Overview 2/2 Read alignment objects Iterations loop (user-defined) File CDB Fit tracks Loop over volumes (user-defined) Align volume(s) CDB Update alignment objects Minimize residuals
Alignment of volume(s) n Base method for aligning volumes: Alignment. Tracks: : Align. Volumes() n What does it do? n n n It aligns a volume A (set of volumes) w. r. t to another volume B (set of volumes) Load space-points arrays with >=1 point in volume(s) A Apply accumulated alignment info (Ali. Align. Obj) for all space-points in volume(s) A and B Fit space-point arrays (tracks) in volume(s) B and extrapolate them to volume(s) A The input is: two arrays (A&B) of ints Arrays with track (volume unique IDs) The output is: updated alignment info for extrapol. points in volume(s) A the volume(s) A n Note: volume sets A and B can (partially) overlap n Several predefined methods to align single volumes, layers are implemented Arrays with all space-points in volume(s) A Calculate and minimize residuals in volume(s) A Update alignment info (Ali. Align. Obj)
Track fitters n n Base class for track fitters – Ali. Track. Fitter: n Interface to space-point array being fitted n Interface for getting the two space-points arrays (residuals) n Abstract Fit() method: n Fits the track within user-defined volume(s) n Prepare the arrays with residuals n To do: all fitters share some part of Fit() method move Fit() to the base class and define some methods inside as abstract n Getters for fit quality information Current status n n Ali. Track. Rieman. Fitter implemented Ongoing development (MI and Cvetan) n Interface to the ROOT TLinear. Fitter (Possibility to use “Robust” fitter) n n Linear fit, parabolic fit, Rieman fit with tilting angles ( for TRD), parabolic fit with tilting angles Interface to the Kalman fitter (Ali. External. Track. Param)
Track Residuals minimization n n Base class for residuals minimization – Ali. Track. Residuals: Two classes implemented: n n Minuit based (Ali. Track. Residuals. Chi 2) Fast linear minimization (Ali. Track. Residuals. Fast): n n n Assume small mis-alignment rotation angles: linear transformation Sufficient precision assuming angles ~mrad Interface to the TLinear. Fitter to be implemented n n Possibility of fixing parameters Robust fit
Alignment - status n n The misalignment implemented in the simulation The correction for the misalignment implemented in the reconstruction n Test with misalignment on the level +-1. 5 mm and angular misalignment 0. 6 degree made The performance of tracking with perfect alignment parameters – almost the same as with ideal geometry First attempts to use alignment framework (“real MC” data) – work in progress
- Slides: 39