Ba. Bar Si Tracker Alignment David Nathan Brown, LBNL Representing the Ba. Bar SVT alignment group l. The Ba. Bar experiment l. The Ba. Bar Si Tracker alignment procedure l. Alignment procedure validation l. Results l. Lessons learned
PEP-II and Ba. Bar 1. 5 T solenoid DIRC (PID) 144 quartz bars 11000 PMs EMC 6580 Cs. I(Tl) crystals e+ (3. 1 Ge. V) Drift Chamber 40 layers 1/3 axial, 2/3 u+v stereo e- (9 Ge. V) Silicon Vertex Tracker 5 layers, double sided strips Instrumented Flux Return iron / RPCs (muon / neutral hadrons) David Brown 2 LHC Detector Alignment Workshop Sept. 4, 2006
Track Momentum on the (4 S) <Pt> ~ 500 Me. V l Scattering (material) largely dominates over point (hit) resolution in impact parameter resolution David Brown 3 LHC Detector Alignment Workshop Sept. 4, 2006
Ba. Bar Physics Goals l Observe CP violation in B system F Time-dependent mixing (e. g. sin 2 ) F z ~ 260 m, zvertex ~ 180 m, 20 m point resolution l PDG-competitive measurement of B, lifetimes F Control average alignment systematics to ~ 1 m (0. 5%) l No Bs mixing, tertiary charm vertex separation, … F Modest requirements on material, resolution David Brown 4 LHC Detector Alignment Workshop Sept. 4, 2006
Ba. Bar SVT l 5 layers, 340 wafers F Radii from 3. 3 to 15 cm F ‘Lampshades’ in layers 4 + 5 l Double sided readout F 90 strips F Kapton fanouts in active region l ~2% X 0 total at normal F 1% X 0 Be beampipe l No hardware alignment David Brown 5 LHC Detector Alignment Workshop Sept. 4, 2006
Wafer Alignment Description Geometric midplane w=0 l Sensor local coordinates F u≈ v≈beam, w≈radial outward l 6 alignment parameters Si Sensor u F Deviation WRT nominal F 3 translations u w v F 3 (small) rotations u w v w u v (0, 0, 0) l Total system has 6 redundant Global alignment DOFs l Internal DOFs F Charge drift asymmetry (=0) F Lorentz shift (estimated) F Non-planar distortions - ++ -- + - ++ - -+ + David Brown ~280 m 6 LHC Detector Alignment Workshop Sept. 4, 2006
Residual min. distance from track to hit in space Track Residuals ~20 m Reduced residual (cm) l Reduced residual excludes the hit from the track fit ~1. 0 Pull = r/ r David Brown 7 LHC Detector Alignment Workshop Sept. 4, 2006
Ba. Bar Alignment History l Ba. Bar design and construction: 1995 1999 F Alignment is considered (overlaps) but not studied l First data and commissioning in 1999 F Used Optical Survey wafer alignment + cosmics l 1 st Alignment procedure development 1999 2000 F Based on (primarily) e+e- + - events F 1. 5 FTE for development and operation F Procedure was manpower, cpu and data intensive H ~1 month turnaround time F Visible systematic errors remained H Early Ba. Bar physics results were not compromised! l Complete rewrite of alignment procedure 2001 2002 F 3 FTE development effort over 1 year F Separate operations effort of 0. 5 FTE F Designed coherently with a new Ba. Bar Data Model F Deployed in 2002, we are still using this procedure today David Brown 8 LHC Detector Alignment Workshop Sept. 4, 2006
Ba. Bar lifetime in year 2000 alignment! variation is ~10% of lifetime Average 1 -3 decay distance Black=data, red=MC David Brown 9 LHC Detector Alignment Workshop Sept. 4, 2006
Alignment Design Principles l Combine complementary constraints F Use lots of tracks to cover all wafer DOFs F Use different event triggers and track geometries to balance systematic biases F Relate wafers across the detector to control global distortions F Incorporate lab-based optical survey information l Select data to provide uniform constraints F Make detector coverage more uniform F Select events uniformly over (short) time period F Equilibrate statistical errors F Minimize statistical correlations between wafers David Brown 10 LHC Detector Alignment Workshop Sept. 4, 2006
Global Distortions l Small relative changes between adjacent wafers that add up coherently across the detector F Residuals work ‘locally’ l Can introduce significant physics bias l Choose alignment constraints which control these R Z R Radial expansion (distance scale) Curl (charge asymmetry) Telescope (COM boost) Elliptical (vertex mass) Clamshell (vertex displacement) Skew (COM energy) Z Bowing (COM energy) Twist (CP violation) Z expansion (distance scale) David Brown 11 LHC Detector Alignment Workshop Sept. 4, 2006
~ few mm Overlaps l Active Si overlap between adjacent wafers in the same layer l Small gap between overlapping wafers F Constrains adjacent wafers F Not as effective in hex geometry l Overlaps cumulatively provide a circumference constraint F Relies on precise knowledge of wafer size F Constrains radial expansion, clamshell distortions l Small fraction of tracks F Between 1% and 3% David Brown 12 LHC Detector Alignment Workshop Sept. 4, 2006
Cosmic Rays l l High-momentum tracks (> 1 Gev) Relates opposite side wafers constrains telescope distortion Off-axis constrains twist, elliptical distortions Low rate, non-uniform illumination David Brown 13 LHC Detector Alignment Workshop Sept. 4, 2006
Pair Fit l Fit 2 tracks from e+e- + - (and e+e-) simultaneously F Constrained to a common origin F Constrain momentum to ‘known’ CM 4 -momentum H Scale errors for beam uncertainties H Implemented in the Ba. Bar Kalman track fit l Provides pair-constrained residuals F Not just a mass-constrained vertex fit! l Constrains curl, bowing, and skew distortions l Technique can work for other track pairs (ie + -) l Depends on initial beam parameter knowledge David Brown 14 LHC Detector Alignment Workshop Sept. 4, 2006
Optical Survey l Use combination of Module Survey (lab bench) + Assembly Survey l Constraint of wafers within a module complementary to tracks l Constrains Z expansion distortion Top ( -z) view (distortions X 50) side (r-z) view David Brown 15 LHC Detector Alignment Workshop Sept. 4, 2006
Survey Constraint l Compute ‘survey to current’ transform using reference wafers F Minimize difference between position of fiducials on the wafers l Predict position of ‘test’ wafer position in ‘current’ alignment l Compute 2 = difference between current and survey position F Multiply out-of-plane errors X 10 to accommodate motion since survey l Add survey 2 to track residual 2 Test Expected position David Brown Test Reference wafers Survey alignment Reference wafers ‘Current’ alignment 16 LHC Detector Alignment Workshop Sept. 4, 2006
Outer Tracking Constraint l Tracks are split at boundary F Each half fit separately l Outer track fit used to constrain the inner track fit Kalman Fit F Can select which parameters to propagate F Improves precision while controlling propagation of outer tracker systematics F Standard feature of Ba. Bar Kalman track fit (d 0, 0, , z 0, tan ) l -pair + cosmic (high p) F Constrain only curvature l Isolated high-P hadrons F Constrained to full outer track fit (5 parameters) l Keeps relative (global) alignment from drifting David Brown 17 LHC Detector Alignment Workshop Sept. 4, 2006
Alignment Data Reduction Central reconstruction Ba. Bar Calibration stream Data 1% of all events Calib. Data Alignment skim 0. 1% of calib. data Align Data l A dedicated sample is selected during reconstruction F pairs, cosmics, prescaled hadronic events with high P tracks, … F Written to a dedicated stream (file) l From ~ 2 days accumulation we extract an alignment sample F Events are prescaled by type and polar angle coverage H Timescale driven by cosmics F Only selected tracks are kept, all other data is removed H Outer tracker info is kept as a fit constraint, reduces track size by 1/3 F Hits are prescaled for uniform coverage, selected hits are flagged H Defines fixed selection of hits used across iterations H Greatly reduces statistical correlation between wafers l Customizations are built in to the Ba. Bar Data Model David Brown 18 LHC Detector Alignment Workshop Sept. 4, 2006
Alignment Iteration l Iteration factorizes the alignment problem F No need for huge matrix inversion (6 X 6 vs 1440 X 1440) F No need to compute distant derivatives l 1 iteration = loop over all wafers l Minimize 2 (closed form) for each wafer F Sum 2 + associated derivatives wrt alignment parameters F Solve for the change in this wafers alignment parameters l Wafer positions are updated only after a full iteration F Parallelizable (if wall-clock time were an issue) l Initialize using previous, survey, nominal, test configuration, … l Tighten residual cuts after partial convergence F Reduces the effect of outliers without biasing alignment F Requires re-writing alignment dataset (reflagging hits) l Convergence when wafers stop moving F p 2 ( P/ P)2/6 < 0. 01 for every wafer in 1 iteration F ~100 iterations, <24 hours real-time (single processor) David Brown 19 LHC Detector Alignment Workshop Sept. 4, 2006
Alignment Convergence Tight residual cuts applied Convergence p 2 < 0. 01 for every wafer (~100 iterations) David Brown 20 LHC Detector Alignment Workshop Sept. 4, 2006
Alignment Operations l Alignment computed every 2 weeks (or as necessary) F Fully automated (except validation!) F 2 -day turnaround F Upload to database only if changes are significant (by a human) l So far we have ~40 alignment periods, separated by F Detector interventions F Humidity effects H Carbon fiber is hygroscopic l Detector has been stable for the past ~2 years History of outer layer relative radial position vs Z for 2001 2003 David Brown http: //dnbmac 3. lbl. gov/~brownd/alignm ent/Svt. Change_dr 21 LHC Detector Alignment Workshop Sept. 4, 2006
Global Distortion Tests l Validate the procedure against global distortions F Small, coherent relative wafer displacement l Use undistorted MC sample composed as data F Cosmics, -pairs, hadronic decays, … l Align starting with a distorted initial condition F 50 m scale, smooth dependence on either R, , or Z R Radial expansion (distance scale) Curl (charge asymmetry) Telescope (COM boost) Elliptical (vertex mass) Clamshell (vertex displacement) Skew (COM energy) Z Bowing (COM energy) Twist (CP violation) Z expansion (distance scale) David Brown 22 LHC Detector Alignment Workshop Sept. 4, 2006
Example: Elliptical Distortion Apply 0. 1% elliptical distortion (~50 m amplitude in layer 5) 100 Iterations before 23 Layer 4 cm Layer 5 ~50 m amplitude David Brown Layer 3 cm cm Layer 5 Layer 2 cm cm Layer 4 cm cm Layer 3 Layer 1 Layer 2 cm Layer 1 cm cm R vs by layer LHC Detector Alignment Workshop after Residual amplitude <5 m Sept. 4, 2006
Z Scale Validation Mat. Int. location mm Tracks from material interactions agree with bench measurements to 0. 03 ± 0. 05 % David Brown mm mm 24 LHC Detector Alignment Workshop Sept. 4, 2006
-pair miss distance d 0 =0 After alignment, we observed a strong 6 -fold symmetry David Brown d 0 (cm) Year 2002 data! 25 LHC Detector Alignment Workshop Sept. 4, 2006
u (cm) v (cm) The Explanation: Wafer Bowing u u l Fit wafer sagitta wafer F Use both u and v residuals F Iterate with normal alignment F Mostly affects layers 1, 2 + 3 Sagitta incident track David Brown w u 26 l Correct in reconstruction u F Model v strips as 3 linear pieces LHC Detector Alignment Workshop Sept. 4, 2006
Wafers are not planes (or cylinders)! 3 -D Interferometric survey of 1 module before installation David Brown 27 LHC Detector Alignment Workshop Sept. 4, 2006
-pair Miss Distance l Average variation of <2 m in d 0, <10 m in z 0 l With 10 X standard alignment sample, structure is seen F More general non-planar distortions cm cm tan David Brown tan 28 LHC Detector Alignment Workshop Sept. 4, 2006
Lifetime Revisited (2005) l “The peak to peak variation of the reconstructed decay length vs is consistent with just natural lifetime fluctuations. ” A. Lusiani final alignment Average 1 -3 decay distance David Brown 29 LHC Detector Alignment Workshop Sept. 4, 2006
Ba. Bar’s sin 2 History Alignment Development David Brown 30 LHC Detector Alignment Workshop Sept. 4, 2006
Si Alignment Lessons Learned l Detector Design F Prioritize material, resolution, stability F Simulate alignment to optimize overlap, layer coverage, … l Construction F Make Lab-bench measurements of all components H Survey aggregate sensor units (module, ladder, …) in 3 -D H Measure material properties of all active-region components V Si thickness, material of hybrids, location of masking, … F Assembly survey as a cross check (if practical) l Software Design F Data model support for alignment H Custom event selection, hit flagging, parameter constraints F Kalman track fit alignment-specific features H Pair fit, parameter constraint F Allocate adequate manpower to alignment development l Operations F Allocate dedicated processing and storage for alignment David Brown 31 LHC Detector Alignment Workshop Sept. 4, 2006
Lessons Learned (continued) l Procedure F Accurately represents the true DOFs H Consider non-planar distortions! F Use complementary event types and external constraints F Prescale events to create a uniform, consistent data sample F Prescale and flag hits H Reduce statistical correlations H Consistent and stable 2 calculations F Validate against realistic distortion scenarios F Don’t get hung up on mathematical details H Any well-behaved, additive measure will probably work H Any minimization technique that converges will probably work l Physics Use F Plan for providing an early (preliminary) alignment F Provide analysts with a misalignment estimate l Be prepared for the unexpected! David Brown 32 LHC Detector Alignment Workshop Sept. 4, 2006
Backup Slides David Brown 33 LHC Detector Alignment Workshop Sept. 4, 2006
How Well To Align? Vertex resolution L Momentum resolution l Statistical (< 5% from alignment) F in-plane < x/3 F out-of-plane ~ in-plane/ l Systematic (no visible biases) F Roughly 3 -times better than statistical on average David Brown 34 LHC Detector Alignment Workshop Sept. 4, 2006
David Brown Pair Fit l Curvature resolution improves >2 orders of magnitude! l Constrains relative dip angle (through boost) Single Fit Pair Fit Results 35 LHC Detector Alignment Workshop Sept. 4, 2006
Lab Assembly Survey Comparison l Compare at fiducials F Remove global DOFs l <3 m in plane F ~1 m statistical l ~20 m out of plane F ~10 m statistical l Average these when used in alignment Lampshade wafers David Brown 36 LHC Detector Alignment Workshop Sept. 4, 2006
Event and Hit Prescaling normal l Prescale events by category -pair F + -, cosmic, overlap track, … l Prescale hits on each track cosmics F Uniformly populate wafers F Sample data period uniformly F Balance different event types F Eliminate statistical correlation between wafers overlaps l Flag selected hits F The exact same hits are used to calculate 2 every iteration F Can (anti-)select hits when validating F Written into the data l Overlaps are under-populated F 1. 5% nominal overlap in layer 4 David Brown 37 LHC Detector Alignment Workshop Sept. 4, 2006
Iteration Control l Iteration is controled by tcl scripts with tk window F Parameters can be adjusted F Job progress is monitored l Typical job converges in ~100 iterations and takes ~ 24 hours David Brown 38 LHC Detector Alignment Workshop Sept. 4, 2006
-pairs after Curvature Correction l Average distortion reduced to ~2 m in d 0, ~10 m in z 0 l With 10 X data, structure is seen! David Brown d 0 39 LHC Detector Alignment Workshop z 0 Sept. 4, 2006
Residual (cm) Aleph VDET bonding error Track position on wafer (cm) David Brown 40 LHC Detector Alignment Workshop Sept. 4, 2006
Скачать презентацию Ba Bar Si Tracker Alignment David Nathan Brown
8b73cec2665ae720ca78222241e3a4d4.ppt