TkrRecon Status Perugia Software Workshop May, 2003

Alignment – 1 st Try: Towers Finding constants (Hiro) –One (central) tower –5M events, 50K cross into the tower –6 parameters, 3 shifts, 3 rots –Root macro, using clusters –Errors on found parameters Shifts: ~ a few microns Rotations: ~ a few microradians, ~ ½ arcsecond –Depends on careful track selection, and using correct track errors –Next: edge towers, more than one tower at a time Misalignment in Recon (Leon) –Incorporated into TkrDigi Interactions in silicon not properly handled yet –Ready to incorporate into patrec/fitting A bit complicated, part way there But! –Leon and Hiro don’t agree! –Shifts and rotation around z seem ok –Rotations in around x an y axes seem to be different –Working on reconciliation

Calibration: Failure modes Two functions –Account for actual dead layers/towers –Allow for killing of layers/towers in simulation, for studies Gleam input –List of failed towers and/or planes in the jobOptions –Later in calibration database, generated from badstrips calibration Digitization –Any hits from these planes are ignored in TkrDigi –For real data, the hits won’t be there anyway Mitigation in TkrRecon –If there is no hit where expected, checks to see if plane is “failed” –If so, carries on, just as if track went through a gap.

Calibration: Bad Strips Main use is tagging actual dead/hot strips Calibration root macro –produces xml file entered for calibration database –Dead and hot strips currently come from same data sample –Scaling? Gleam input - either/or: –Automatic from database Choice based on timestamp, not yet available –File of strip lists in jobOptions Digitization –Default is no special action –Kill flag in TkrDigi jobOptions Clustering –Ignores contiguous isolated bad strips –Adjacent clusters connected by bad strips are merged Mitigation in TkrRecon –None yet –Plan: check for bad clusters in case of a missing hits. Bad cluster should be: Close enough Wide enough –Requires reworking of clustering code to make “pure bad” clusters

Calibration: ToT ToT calibration is at a very primitive stage. Current algorithm: –Looks at entire half-plane –Fits ToT distribution to Landau. –Estimates lower edge (measure of efficiency) –Very simple cuts! Plan –Add quality information –Look at each chip separately? 13,824 in flight instrument! –Better cuts Gleam –Just carries information around, not used yet EM data

Digitization Two Digitization packages TkrSimpleDigi –Simple! –Bells and whistles added Alignment, Failure modes, relational tables, etc. TkrBariDigi –Very detailed simulation of physical processes: Ionization in silicon Drift Charge collection vs time Electronics response –Slow –Has fallen behind on B&W Digi-Merge Project (Michael) –One driver algorithm –Break up code into functional modules –Common code for common functionality

Geometry – Some Recent Discoveries Radiator was incorrect for SuperGlast layers. –Thickness was adjusted for W_alloy material 92.5% W, 5% Ni, 2.5% Fe, ρ = 17.6 –Material was still pure W Position of TKR/CAL was incorrect –Both were ~ 3 cm too high –Have been approximately corrected, but not really right Geometry review planned for mid-June. –Subsystems have agreed to supply official numbers

Iterative Tracker Recon What is it? –The Iterative recon is a mechanism for allowing parts of the Tracker Reconstruction software to be called more than once per event. –In particular, existing pattern recognition tracks can be refit and the vertex algorithm rerun. Why is it needed? –The Calorimeter needs/wants the output of TkrRecon in order to refine the initial energy determination. –The TkrRecon wants the best energy estimate in order to get the best track fits and, subsequently, the best vertices. –The Iterative Recon solves this problem by providing Cal Recon with tracking information to get an improved energy for an event which is then used to refit the tracks and rerun the vertexing. –This process can be repeated as many times as the user likes.

Iterative Tracker Recon Basic idea: –Find the CalClusters and current Pattern/Fit Tracks in the TDS –Reassign Calorimeter energy to the current set of Fit Tracks Algorithm is currently identical to that used in the Combo Pat Rec Only assigns energy to the two best tracks… –Rerun the vertexing Two Assumptions: –The results of the Second Pass CalRecon (ie the improved energy determination) are obtained from the CalClusters via the method “getEnergyCorrected” –Tkr Clustering and Pattern Recognition do not need to be rerun.

Iterative Tracker Recon Does it work? –Iterative Recon works –Need to better understand the 2 nd pass energy recon! “WAgammas” –1 GeV –5° cone about normal –Into 6 m 2 area containing Glast Red Histograms –Energy of the “best” (first) vertex Blue Histograms –Energy of the two “best” tracks (if two or more tracks) Note Energy shift in 2 nd pass! First Pass Recon 2 nd Pass Recon

List of TkrRecon Topics to Discuss here