A High-Rate TPC for PANDA Christian Höppner Technische Universität München C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Outline The PANDA Experiment The PANDA TPC Space Charge Effects Tracking of Cosmic Myons with a small GEM-TPC Pattern Recognition A Generic Kalman Filter Implementation for Global Tracking in PANDA C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA: Antiproton Annihilations at Darmstadt Facility for Antiproton and Ion Research (FAIR), GSI, Darmstadt High Energy Storage Ring: 1 – 15 GeV antiproton-beam Continuous beam Internal p-target Hadron-physics: charm-spektroscopy, gluonic excitations, hypernuclei, …. C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

The PANDA Spectrometer 2 x 107 events / s ~ 3 tracks / Event 4π - spectrometer: 2T solenoid-field Antiproton Beam C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

The PANDA Spectrometer Central Tracker Momentum measurement: dp/p ~ 1% Material budget < 1% X0 PID via dE/dx Proposal: Highrate-TPC with continuous readout Also proposed: STT Decision for PANDA TDR 2009 C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Geometry: Radius: 15-42 cm, Length: 1.5 m C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Electron drifttime ~ 50 μs (at 107 Events / s) Event Mixing (500 Events happen per e--Drifttime) Although ion backflow is suppressed: Some ions make it to the drift volume → Accumulation of space charge Geometry: Radius: 15-42 cm, Length: 1.5 m Challanges for high-rate TPC C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Simulation of Space Charge Realistic GEANT Simulation Ion-backflow ε=1 (1 ion from amplification per incoming primary electron) realistic values 1 < ε < 10 Ion-backdrift: 1.7 cm/ms Equilibrium reached after 100 ms QUESTION: What’s the effect of this space charge on the reconstruction of tracks? Two Contributions: Primary ionization & ions from amplicfication Electron drift Amplification & Readout Beam C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Track Distortions due to Space Charge Displacement of several mm possible Mean: ε = 1 1.5 mm ε = 10 5.2 mm Has to be corrected for in reconstruction (as done e.g. in ALEPH) Difference of reconstructed and generated electron starting point ΔR [cm] R [cm] z [cm] Also, slightly imperfect solenoid field has been taken into account C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 GEM – TPC Test Chamber cosmic myon Scintillator + PMT Triple GEM - + + - 80 mm + - - + + - Ar/CO2 + - Elektronics: ALICE TPC – Inverter/PASA/ALTRO (noise ~ 1900 e-) C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 GEM – TPC Test Chamber Elektronics: ALICE TPC – Inverter/PASA/ALTRO (noise ~ 1900 e-) C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Event Topology z primary electron clusters Investigate tracks which are rather perpendicular to readout plane → important topology for PANDA y x C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Event Topology z y x C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Event Topology Per Pad: puls finding (MultiFit) z A y t0 x C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Event Topology z y x C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Event Topology z y With samples (t0,A): Clustering in x-y x C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Event Display Schematic Data z y x pattern recognition: hough transform fit: chi2-minimization C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Event Display Schematic Data z y x pattern recognition: hough transform fit: chi2-minimization C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Pattern Recognition – Hough Tranform Straight Lines: E.g. in yz-plane: y = a z + b each point (yi, zi) becomes a straight line in the parameter space (a,b) b = (-zi) a + yi the lines in (a,b) space of collinear points intersect in a point C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Pattern Recognition – Hough Tranform Straight Lines: E.g. in yz-plane: y = a z + b each point (yi, zi) becomes a straight line in the parameter space (a,b) b = (-zi) a + yi the lines in (a,b) space of collinear points intersect in a point C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Pattern Recognition – Hough Tranform C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Pattern Recognition – Hough Tranform C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Pattern Recognition – Hough Tranform Also applicable to helical tracks (if they intersect the origin [x,y]=[0,0] ) Pattern recognition is done completely in x,y plane R is distance of center of circle θ is azimuthal angle of track in origin In image space (R, θ) for each point (xi,yi) a curve is drawn: The curves from points on the same helix (R,θ) intersect in one point. C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Large PANDA-TPC prototype A prototype with 50 cm drift length and 30 cm diameter will be built before the PANDA TDR in 2009 Detector will be used in FOPI Study materials for construction of final detector mechanical stability gas purity HV insulation Study HV supply to field cage insulation resistor chain stability Study mechanical properties of field cage and readout plane Study integration of front-end electronics Study performance with close-to-final front-end electronics Study distortions, calibration, corrections C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Large PANDA-TPC prototype Electronics: N-XYTER asic looks promising (GSI detector lab for CBM) analog zero suppression Testing pending C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

A Generic Kalman Filter Implementation C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Objective: seperate algorithm from parametrizations Kalman filter algorithm Pure linear algebra Parametrizations Hits: examples: TPC (r, φ, z) STT (x,y,dr) PixelMVD (x,y,z) Tracks: examples: Barrel Spectrometer: some helix-representation Forward Spectrometer: e.g. local straight line parametrization C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Class Structure of Generic Kalman Filter User implements: concreteHit concreteTrackRep C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Kalman Filter Ingredients As formulated in Frühwirth et al. “Data Analysis techniques …“ C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Distribution of Responsibilities All information of hits up k-1 is already in the state vector and the covariance matrix xk-1 and Ck-1 → Process Hit k Kalman Kalman Gain Update Track propagation / prediction TrackRep Calculation of residuals Hit C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Distribution of Responsibilities Kalman Kalman Gain Update Track propagation / prediction TrackRep Calculation of residuals extrapolate to where? Hit C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Distribution of Responsibilities Kalman Kalman Gain Update Track propagation / prediction TrackRep Calculation of residuals Hit C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Distribution of Responsibilities Voila! Updated state and covariance! Next iteration could without any problem use hit from different detector that measures very different coordinates Kalman Kalman Gain Update Track propagation / prediction TrackRep Calculation of residuals Hit C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Example Straight Line Fit trackRep Free parameter: z State vector: xk=(ax,bx,ay,by)T Extrapolation (trivial) HIT (some pixel detector) Free Parameter: z Hit coordinates mk = (x, y)T H Matrix: Residual C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Example: Helix Fit trackRep Free parameter: r State vector: xk=(Φk,zk,λ,C,D)T Extrapolation Covariance matrix prediction: Calculation of Jacobian matrix numerically or analytically HIT (e.g. TPC) Free Parameter: r Hit coordinates mk = (rΦ, z)T H Matrix: C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Example: Helix Fit Circles: hits red dashed: start value for fitter green: result of fit C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Example: Helix Fit Off-diagonal elements of covariance matrix a very large for representation → we need a better suited track representation xk=(Φk,zk,λ,C,D)T Circles: hits red dashed: start value for fitter green: result of fit C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Example: Local Straight Line Fit trackRep 6th order Runge-Kutta solver for propagation in arbitrary magnetic field Numerical jacobi matrix calculation for error propagation Solenoid Dipole z xk = (xk,yk,xk‘,yk‘,q/p) Circles: hits red dashed: start value for fitter green: result of fit C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Idea: Generic Kalman with GEANE trackRep Free parameter (length along track) and track representation (state vector) defined in GEANE way Extrapolation of track and error propagation including multiple scattering and energy loss done by GEANE HIT (for each detector kind) User provides H-Matrix for transformation from hit coordinates to GEANE track parameter space That’s it! Code is basically ready to be used with GEANE C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Global Track Fit in PANDA C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Challanges for the PANDA TPC Project Space Charge Event Mixing Primary reactions at 107 / s with ~ 3 tracks / event Electron drift time is ~ 50 μm → tracks of 500 events are in the chamber at a given time Continuous operation → tracks can a priori only be reconstructed with unknown offset in z Finding an absolute time for event a track belongs to: - endcap penetration or target pointing can help - final z-pinning with hits from other detectors (vertex tracker) C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Concepts for continuous readout at high rates What could it look like? Raw data from ADC cards: 500 GB / s (times and amplitudes) To computer farm: ~ factor 10 less (clustering, tracklet recognition, ..) Some online track fitting in computer farm Matching with other detectors or other means of finding an absolute time for the event Construction of events Physics selection (trigger) - Simulations are being done on hardware clustering implementation - Track reconstruction algorithms are being developed C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

Conclusion and Outlook GEM-based TPC proposed as central tracker for PANDA Space charge effects will distort tracks in order of mm → has to be corrected for in reconstruction 8 cm test-chamber is running nicely – ALICE electronics seems to be under control → study time structure of signals and spatial resolution (of perpendicular tracks) First experiences with Hough transform has been gained with test chamber → implement cylindrical tracks for PANDA software A generic implementation of the Kalman Filter has been used for first track fits in the PANDA software framework → searching for best track representations ?? A large prototype will be built before the PANDA TDR in 2009 – this detector will be used for a new physics program at FOPI C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007

PANDA-TPC TPC JAMBOREE Aachen March 2007 Fit Results C. Höppner (TUM E18) PANDA-TPC TPC JAMBOREE Aachen March 2007