Run II Reconstruction Program Status - Version p13

Overview

The p13 version of the reconstruction program represents a major milestone for the DØ experiment. It was originally intended to be the version used to reconstruct data taken after the "October, 2002" accelerator shutdown. However, that shutdown was postponed, and instead efforts have been refocused towards reconstructing "50 pb-1" for the Spring, 2003 conferences (Moriond, etc.)

One of the primary improvements targeted for p13 was to deploy the best currently available set of track reconstruction algorithms. A task force was formed to review the performance of existing algorithms, and recommended that RECO use as default the "GTR + HTF" algorithm. Some details concerning this review are available.

In addition, the various Algorithm and Object ID groups had the following goals for p13:

• Detectors / Alignment / Calibration
  • Improve performance (memory usage) for calibration database access.
  • Align calorimeter with tracker.
  • Support new muon Monte Carlo geometry (p12+ MC samples).
  • Improve SMT and CFT alignment
  • Split unphysical CFT clusters.
  • FPS unpacking with correct readout map.
• Tracking
  • Implement TARC recommendation (GTR + HTF).
  • Provide simple interface to select different algorithms from the command line.
• Vertexing
  • The ability to refit a vertex using the thumbnail.
• EM ID
  • Deal with aligned calorimeter.
• MU ID
  • Understand and fix local muon track fit failures.
  • Clean up muon tracks sharing segments to avoid duplication.
  • Certify muon-calorimeter matching.
  • Optimize and certify track matching to limit double counting in overlap region.
• Tau ID
  • Deploy mulitple neural networks.
• Jets/MET
  • Settle issues raised by calorimeter task force.
• B ID
  • Incorporate tagging improvements into RECO as developed with reco_analyze ROOT tuple.
  • Enable redoing B-tagging using thumbnail
• FPD (not yet integrated in RECO; expected for p14)
  • First release of the fpd_reco framework package, does track reconstruction on FPD detectors in multi-track environment.
  • First release of fpd_unpack package.
  • Unpack data from one AFE board (Dipole spectrometer)
  • Fixes for gcc compliance
  • Fixed FPDDigiChunk definition
  • FPDTrackChunk defined
• Data Tiers
  • Physics ready thumbnail (TMB) with size 10 KB.
  • Physics ready DST with size 150 KB.
• General
  • Implement ZERO BIAS streaming to support Monte Carlo - write out Raw Data Chunks for all ZERO BIAS triggers to a separate "stream".
  • Disable "special streams" that were previously used to tag "interesting physics events".
  • Drop support for production running of d0_analyze / reco_analyze.
  • Perform all tests with floating point exceptions enabled.

Based on requests from the WZ and Top physics groups, the following goal was also established:

• Produce a TMB+ that includes the CalDataChunk. This will require an estimated additional 3 TB of SAM cache to support. This request will be honored until February, 2003, when the standard TMB will be restored.

Status

Now...

p13.06.01 is running on the data farms. p13.07.00 is running on the MC farms.

History...

According to the "Overall Goals of the Experiment", p13 RECO had two major milestones - cut the first version the week of September 30 and make it available for farm production the week of October 28.. On Friday, October 4, we started our final build of t02.35.00 which was copied to make p13.00.00, achieving the first p13 milestone. On Friday, November 1, the final build of p13.02.00 completed. It was frozen on November 4 and began official production of data on November 5.

This successful achievement of the p13 RECO goals was announced in D0 News.

One of the highest profile goals for RECO p13 was the implementation of the recommendations of the TARC committee regarding the default tracking algorithm (GTR + HTF). This has been accomplished and preliminary studies indicate performance consistent with TARC p11 testing. As a cross check, p11 GTR-only results were also reproduced. Other general performance numbers (average number of muons and electrons (tight, medium, loose), mean and RMS MEX, MEY and MET, CPU time, memory, DST size, TMB+ size) look reasonable. More information on RECO performance is available.

The following major goals have been achieved:

• GTR + HTF tracking algorithm successfully deployed.
• CFT axial-only tracks saved.
• Command line interface to select tracking algorithms implemented.
• Correct FPS unpacking.
• Clean up muon segments.
• Loose muons in thumbnail.
• General muon id improvements.
• Fix calorimeter offline threshold "bug" so offline is same as online.
• Install new calorimeter noise file.
• Multiple tau neural networks.
• Debug thumbnail.
  • Status of thumbnail as of October 30, 2002
• Streaming of zero bias events.
• Support streaming dilepton events.
• "Special stream" disabled.
• DST and thumbnail size goals reached.
• TMB+ deployed to support physics group requests.
• Fix many floating point errors.

In addition to RECO, there are developments in the Level 3 system. The status for p13 is available.

The following RECO test samples are available:

• p13.08.00
  • Monte Carlo RECO certification samples
• p13.05.00
  • Monte Carlo RECO certification samples
• p13.03.00
  • Dielectrons (data)
  • Monte Carlo RECO certification samples - Problems have been reported with these samples. Too much noise was included in the calorimeter.
• p13.02.00
  • Monte Carlo RECO certification samples
• p13.01.00
  • Dielectrons (data)
• p13.00.00
  • Z(ee) data
  • Z(mumu) data
  • Monte Carlo RECO certification samples

Other collections of interesting event samples:

• EM ID event samples
• MU ID event samples
• B ID MC event samples
• JES event samples
• B Physics MC and data samples
• Electroweak data samples
• Higgs Physics MC and data samples
• New Phenomena samples
• Top Physics data samples

Performance

The following numbers indicate some general performance characteristics of RECO p13.02.00. These numbers are representative; we observe fluctuations depending on exactly which events are processed. More recent versions of p13 have similar performance.

Standard plots are also available (p13.00.00):

• Run 155554 - An old run. (You can compare to p11.11.00 results.)
• Run 160588
• Z(ee) data
• Z(mumu) data
• Tracking efficiency in data using dimuon events

Other plots

• Plots from Erich Varnes on using muons to measure tracking efficiency.

Upcoming Features

Next RECO pass release: None scheduled.

Known Problems

Major error, Annoying feature, Functionality to be added
Scheduled for next release, Problem fixed.
(common = occurs within 100 events, rare = occurs within 1000 events)

Status	Description	Fixed in version
	Bad PDT data taken in run range 168618 to 169290. Not a RECO problem per se, but documented here to help "spread the word".
	Currently available RECO rcp's do not support reprocessing Monte Carlo trigsim output so that l1l2chunk is available in the thumbnail. See "User Information" below for a prescription to get around this problem.
	Analyzing skimmed thumbnails can take a long time. Part of the problem is the current code reloads the magnetic field each time a new run number is encountered. This will be fixed in a future release, however users may make local executables with a patch to cure the problem. In addition, users can customize the unpacking of the thumbnail for their purposes and speed up processing considerably.
	p13.04.00 is taking up to 45 GHz-seconds per event to process certain recent data on the farms. This is about a factor of 2 more than benchmark studies and seems to occur when the instantaneous luminosity is at its highest, resulting in large tails in CFT . For these runs, the HTF tracking time grows very large. The data indicates ~ 1% of the events cause a factor of ~ 2 increase in CPU time required. This appears to be the first evidence of the expected rapid increase in tracking CPU time with multiple interactions. This problem will have to be dealt with in p14.	N/A
	recocert certification statistics don't handle duplicate muon objects correctly.
	Jet reconstruction complains about missing Monte Carlo chunks when processing data. Won't be solved in p13.	N/A
	RECO crashes with floating point errors on d0mino. The same events do not have this problem on Linux. Unfortunately, totalview on d0mino crashes when trying to debug. Users are advised to "unset TRAP_FPE" to get around this problem. Won't be solved in p13.	N/A
	Can't read thumbnails generated with p13.01.00 using p13.02.00 or beyond. This problem will not be fixed.	N/A
	L1cal eta values for calorimeter towers is wrong. A prescription for how to deal with problem is available.	p13.07.00
	Use SMT and CFT calibration databases (currently use flat files). Version p13.07.00 provides this functionality and has been tested using a small number of jobs. More extensive testing will be performed on the farms, and the plan is to move to this mode of operation in p14.	p13.07.00
	An incomplete CFT calibration file was installed on the farms for processing runs 169521 - 170008, resulting in missing stereo hits for one sector. This resulted in poor 3D reconstruction efficiency in that sector. More details are available.	Data has been reprocessed.
	When generating events with p13.05.00 (and previous versions), a small number of cells end up with anomolously large energies. A description of the source of the problem is available from Robert Zitoun.	p13.06.01
	RECO crashes on IRIX using maxopt build, due to uninitialized variable _pshwrE in EmTmbObj.	p13.06.00
	The calorimeter non-linear corrections were not applied in p13.04.00. This results in worse resolution, especially in the EC. The problem is limited to p13.04.00 (i.e. previous versions are ok). A study documented the effects on jets (plots).	p13.05.00
	When reading thumbnails produced with p13.02 using p13.04, the charge of the muon is always zero.	p13.05.00
	The definition of the muon thumbnail variable qptloc is different for p13.02/03 and p13.04. In p13.02/03 it is q*ptloc In p13.04 it is q/ptloc where pt is the local pt of the muon.	p13.05.00
	Memory leak in gtrbase::GTrack::get_num_measurements caused problems with default running of RECO and RecoCert_x. This problem does not show up on the farms, because they disable recocert processing. However, "normal users" are affected.	p13.05.00
	Crashes in p13.03.00 muonid due to overwriting fixed arrays (rare)	p13.04.00
	No FPS clusters in thumbnail.	p13.04.00
	Memory leaks have been identified in the code used to unpack the thumbnail. This affects users trying to analyze p13.02.00 thumbnails, limiting the number of events that can be processed per job.	p13.04.00
	If 2 local muons share hits they are considered as coming from the same physical muon; one needs criteria to chose the BEST. Until now the criteria was using (among other things) the pT of the local muon track. Detailed studies show that this introduces an important bias in pT. This will be changed to use the chi2 instead of the pT. (Christophe Clement)	p13.04.00
	For one local muon track, there are several matchings with different central tracks which are kept into MuonParticleChunk (can deal with that by checking the local track and central track indices).	p13.04.00
	No trigger information in thumbnail.	p13.04.00
	Missing quotation mark in MuoCalTrackReco.rcp causing "email danielw@fnal" error message.	p13.04.00
	The solenoid field is shifted by 1cm in Z in RECO with respect to ideal position. In addition, data indicates possible additional shift. Proposal to fix the known 1cm shift in MC and data in p13.04.00.	p13.04.00
	In p13.02.00, unphysical CFT clusters were split. This seems to result in a 50% increase in processing time, with little change in performance. Proposed to revert back to previous handling of clusters.	p13.04.00
	Thumbnail does not contain information about cells removed by NADA. Thumbnail now includes CalNadaChunk	p13.04.00
	RECO_ANALYZE crashes in vertex_analyze because the quark list in BlockQuark.cpp exceeded the allocated size.	p13.04.00
	RECO crashes with an assert in MRCCtrlMCHModule (pmod->size() != 2). Caused by corrupt raw data. Change assert to ErrorLogger error message.	p13.03.00
	RECO crashes with a segmentation violation in muonid::MuonFind1 (rare). Caused by corrupt muon raw data. Protection to be installed to prevent crashing.	p13.03.00
	RECO crashes with a segmentation violation in cps_reco (rare). Caused by corrupt CPS raw data. Protection to be installed to prevent crashing.	p13.03.00
	RECO crashes in cps_evt/CPS3DCluster with the exception "SLC cache not set" (rare). Seems to be same cause as above problem.	p13.03.00
	RECO goes into an infinite loop in impTagTrackCategory::GetProbability (rare - occurs more often when processing p13 single particle Monte Carlo).	p13.03.00
	RECO crashes with a floating point exception in impTagAlg::calculate() (rare). Protection to be installed to protect against divide by zero.	p13.03.00
	The 8x8 hmatrix chisq in the CC appears to have a problem. The EC looks ok. Reported in w-wenu p13.02.00 recocert Monte Carlo sample. Caused by incorrect noise/zsup file in p13.02.00 d0sim.	p13.03.00
	Bug in calorimeter_geometry affecting extrapolations using CellLine	p13.03.00
	Calorimeter geometry in RECO shifted with respect to p12+	p13.02.00
	Muon geometry in RECO shifted with respect to MC p12+	p13.02.00
	RECO crashes with a divide by zero in ChPartReco (rare).	p13.02.00
	RECO crashes when processing MC event generated with p13.01.00 in gtr_htf::ForwardDetLayer::simHits. Error messages include ? wrong f-wedge width/length = 0.83647, 2.9605, 3.9635 ? cannot get det with layerId = 101, detId = 1.	p13.02.00
	RECO crashes with assert (short_hop || long_hop) in CftPropInteracting::err_dir_prop (rare).	p13.02.00
	RECO crashes in FPSUnp2Digi_data (rare).	p13.02.00
	Calorimeter geometry that aligns it to tracker (3 cm Z shift) causes floating point exceptions (currently disabled).	p13.02.00
	Bug in calculating SMT cluster errors - affects H disks, 1D ladders (dummy n side, error is 144 and not 12).	p13.02.00
	RECO crashes in muonid::MuonFind1 (rare)	p13.02.00
	Inconsistent access to EMQualityInfo methods between DST and TMB.	p13.02.00
	Offline zero suppression algorithm not identical to online ("<" instead of "<=")	p13.01.00
	FPSWedgeData issues an assert when encountering negative ADC values (rare).	p13.01.00
	No FPS clusters (incorrect RCP used by RECO).	p13.01.00
	RECO exits with no error message, due to bug in LinearAlgebra Matrix class (calls exit() if user tries to invert a singular matrix). Need to install protection in vkalman_util/KTrack.	p13.01.00
	RECO crashes with a divide by zero in muo_caltrackreco/MuoCalTrackFinder.	p13.01.00
	RECO crashes with a divide by zero in muon_centralmatch/PropagateMuon.	p13.01.00
	RECO crashes with segmentation fault in muonid::MuonFind1. New version of muonid being tested.	p13.01.00
	RECO crashes with a segmentation fault when accessing CalData. It is currently suspected that there is a memory overwrite someplace in the code. New version of muonid package is being tested.	p13.01.00
	RECO crashes at the end of the job after processing all events. New version of analysis_utilities is being tested.	p13.01.00
	RECO failed to build in p13.01.00 due to change in L3Chunk interface. Fix available in p13-br version of l3fchunk.	p13.01.00
	RECO in p13.01.00 can't connect to run config db server. Fix available in the p13-br versions of config_db_client, mag_field_config, d0omCORBA.	p13.01.00
	RECO crashes at the end after processing all events. Fix available in the p13-br version of recocert.	p13.01.00
	No CPS clusters created.	p13.01.00
	RECO crashes with a floating point error in muo_segmentscintonly::MuonScintLinFit::d0Fit (rare). Temporary user solution - skip offending event.	p13.01.00
	Refitting of tracks eliminates CFT axial-only tracks.	p13.01.00
	Including CFT axial-only tracks causes floating point exceptions.	p13.01.00
	TMB contains Muon::MuoTrackChunk instead of Muon::TrackChunk. This chunk will be obsolute once loose muons are available in the TMB (under development).	p13.01.00
	RECO crashes in em_util::CorrEemcalo::corretada_IC_S with a floating point exception (rare).	p13.01.00
	RECO crashes in d0om writing out MuonParticle (common). Temporary user solution - skip offending event.	p13.01.00
	Including CFT axial-only tracks causes crashes in KTrack due to singular error matrix inversion.	p13.01.00
	Crash in taureco due to axial only tracks.	p13.01.00
	Bug in CFT clusters - not using aligned geometry correctly for stereo clusters.	p13.01.00
	Muon geometry not consistent with p12+ Monte Carlo.	p13.01.00
	CPSClusterChunk takes 32KB in the DST (largest chunk). Will be dropped from the DST.	p13.01.00
	Loose muons in thumbnail.	p13.01.00

Compatibility Summary

The following attempts to describe which versions of p13 RECO are compatible with each other. This summary is very general, and may not cover all possible issues that someone doing an analysis may care about. More detailed information about the differences between versions can be found below. Detailed release notes are also available.

Entries with the same color indicate compatibility. Columns are included if some significant non-compatibility can be identified.

Version History

• t02.35.00 (October 4, 2002)
  • Last development version before cutting production release
• p13.00.00 (October 7, 2002)
  • Not for production use.
• p13.01.00 (October 25, 2002)
  • Not for production use.
• p13.02.00 (October 29, 2002)
  • First official p13 production version for data processing.
• p13.03.00 (September 12, 2002)
  • First official p13 production version for Monte Carlo processing.
  • Equivalent to p13.02.00 for data processing (only contains fixes for crashes).
• p13.04.00 (November 25, 2002)
  • This version is very similar to p13.02.00, but has a few significant changes that make it formally different. Individual users will have to decide if these changes are crucial for their particular analyses:
    • Thumbnail - More information is available in the p13.04.00 thumbnail:
      • L1/L2 information
        • This information allows users to deal with the problem in ScriptRunner.
      • Muon quality words, to help deal with mixed events.
      • Include CalNadaChunk information.
      • FPS cluster data.
    • DST - Include CPSClusterChunk.
    • Improved SMT and CFT alignment. Improvements are for both internal SMT and CFT alignment, and SMT to CFT relative alignment. The shifts with respect to p13.02.00 are small but are formally different.
    • The center of the magnetic field was shifted by 1 cm with respect to p13.02.00. An artificial 1 cm shift was found in the code and was removed. This shift is consistent with analyzing J/psi data.
    • Fix error in the way phi is calculated for certain tracks (wrap around bug) when reconstructing vertices, resulting in an icrease the number of found vertices of 10-20%.
    • Improve dealing with "duplicate" muon candidates. Change uses chisq instead of pT as a criterion.
    • Do not split unphysical CFT clusters. Unlike p11, in p13.02.00 these clusters were split. While it was expected that splitting unphysical clusters would improve tracking efficiency, no improvement was observed. However, small increases in CPU time and memory were observed. It was decided to revert back to the previous handling of these clusters until we understand them better. This change is expected to have a very small effect on tracking performance.
    • Deal with the alignment calorimeter geometry with reconstructing "soft electrons" using semreco.
    • Include energy loss when fitting tracks. This effects very low energy tracks, and the only observed affects are seen when processing Monte Carlo, although the data might improve.
• p13.05.00
  • Fix a bug introduced in p13.04.00 which caused the calorimeter non-linear corrections to not be applied correctly. This bug resulted in worsen calorimeter resolution, especially for the EC.
  • This also version contains two bug fixes that do not affect official production but improve the use by users:
    • Fix memory leak in GTrack:get_num_measurement() which affects running any executable using the recocert package (including RECO).
    • Fix a bug in the handling of SMT 1d clusters so that refit AA tracks can be saved correctly. (AA is not being run in production.)
• p13.06.00
  • Only minor bug fixes not affecting production.
• p13.06.01
  • Contains two important bug fixes:
    • This version is required to process events taken with the p13 version of Level 3. For the same reason, it is also required to process Monte Carlo events run through the p13 version of TRIGSIM.
    • A fix to the calorimeter non-linearity simulation is included in the production version. A description of the problem is available.
• p13.07.00
  • Fix L1/L2 information in thumbnail for Monte Carlo.
• p13.08.00 - No RECO changes

Release Notes

• t02.35.00
  • Test release used as basis for p13.00.00.
• p13.00.00
  • Special fix required for bc_impTagreco/rcp/TrackCategoryDatabase.rcp.
• p13.01.00 (Build 1)
  • Bug fixes
    • bc_eTagreco - Add default variables for lilelihood tag variables, to protect against FPE errors.
    • bc_likeTagreco - Add default variables for lilelihood tag variables, to protect against FPE errors.
    • cft_trfclus - Make stereo CFT clusters correctly use aligned CFT geometry.
    • chpart_evt - Fix FPE errors when trying to keep CFT axial-only tracks.
    • cps_util - Fix CPS clustering code to restore clusters.
    • em_util - EMcandidate : fix EMcandidate to initialize energies in all floors; CorrEemcalo : apply the geometrical energy corrections above 5 GeV and protect log function against negative argument.
    • em_util - Protect CorrEemcalo::corretada_IC_S from a FPE error by establishing a cut-off in the correction.
    • geometry_db - Fix bug in GeometryVersionTable.cpp.
    • gtr_find - Change rcp's to support fix for CFT clusters in cft_trfclus.
    • muo_segmentscintonly - Fix floating point error in muo_segmentscintonly::MuonScintLinFit::d0Fit.
    • recocert - Fix RECO crashes at end of job in recocert.
    • tmb_analyze - Restructure macros for easier build of TMTreeClasses_C.so; protect against JetChunk's with no LeBob objects.
    • vertexreco, vtrkimp, vertex_init, vkalman_find - Require SMT hits when vertexing to avoid crash when CFT axial-only tracks are present.
    • vertexutil, vkalman_util - Get correct B field.
    • vertex_analyze - Turn off V0 block.
  • Functionality improvements
    • analysis_utilities - Support streaming of dilepton events.
    • bc_impTagreco - Use ITrack rather than Vtrack for DCA computation (for redoing tagging off TMB). Updates to track category DB (flexible RCP input, bug fixes, cut variables added).
    • cft_evt2D, gtr_find, trffind - Remove refitting step for CFT axial-only tracks to preserve them if they are saved in the 401 chunk.
    • cft_unpdata, cft_calibdata, config_base - Record which CFT pedestals were used in Global object.
    • d0reco - After fixing cps_util, the CPSClusterChunk takes 35KB/event. This makes the DST larger than its target. In response, the CPSClusterChunk will be dropped from the DST. However, the CPSDigiChunk will be saved, so that CPS clusters can be regenerated as needed for DST analyses.
    • d0reco - Enable dilepton streaming.
    • geometry_db - Install muon geometry to support p12+ Monte Carlo.
    • gtr_find - Reduce z resolution parameter from 1000 cm to 10 cm.
    • muo_segmentlinkedlist - Take the best combination between several WAMUS and FAMUS segments.
    • muo_trackreco, muo_centralmatch, muonid, muo_analyze - Implemente loose muons, MTC matching and isolation variables into muonid and put corresponding variables into the TMB.
    • muo_util, muo_hit, muo_segmentcombi - Implement the PDT dependent time to distance relation.
    • smtutil - Fix SMT hardware map for 2 F disks (for data after run 165894).
    • smtraw - Latest SMT pedestals.
    • tau_nn, tau_evt, taureco, tau_analyze - TauNN class has been updated in order to implement multiple neural networks for tau-identification.
• p13.01.00 (Build 2)
  • Bug fixes
    • l3fchunk - fix const for getPhysToolMap(). This problem appeared in build 1 of p13.01.00.
    • muonid - remove deletion of SurfacePtr, which was causing crashes.
    • muonid - protect against uninitialized pointer.
    • analysis_utilities - remove deletion of D0TriggerSelector singleton, which was causing crashes.
  • Functionality Improvements
    • config_db_client, mag_field_config - Work with new d0omCORBA libraries.
    • d0omCORBA - Work with new db servers.
    • db_server_gen v5_1, db_server_base v2_1 - New db servers.
• p13.01.00 (Build 3)
  • Bug fixes
    • muo_caltrackreco - Protect against divide by zero.
    • muo_centralmatch - Protect against divide by zero.
    • vkalman_util - Protect against inverting singular matrix.
    • tau_nn - Fix clipping of input variables in TauNN.
  • Functionality improvements
    • gtr_find, trfutil, timekeep - Permit CPU timing measurements of individual tracking algorithms.
    • fps_unpdata - Fix FPS unpacking.
    • d0reco - Use correct FPS rcp for unpacking.
    • tau_nn - Update of neural network rcp files; new training for neural networks (one NN for each tautyp).
• p13.01.00 (Build 4)
  • Bug fixes
    • d0reco - Add MuoCalTrackChunk, MuoCentralMatchChunk to list of chunks to be dropped for reprocessing DST's.
    • d0reco - Include RegFPSUnp2Digi_data in OBJECTS file.
  • Functionality improvements
    • calunpdata, caltables - Correct offline zero suppression bug; use new noise and zero suppression file from Robert Zitoun.
    • event_info, config_base - Add support to put stream name into thumbnail.
    • run_d0reco - Install new cal noise file in rundata.
    • muonid, muo_analyze - Add muon isolation variable with a cone of 0.1.
    • muonid - Add independent cut on number of hits for loose muons.
• p13.02.00 (Build 3 - 1&2 were copies of p13.01.00)
  • Bug fixes
    • cft_trfclus - Protect against large, unphysical Z values which cause a crash.
    • fps_unpdata - Protect against overflowing fixed arrays.
    • muonid - Add more protection against empty ChargedParticle chunks. Properly extrapolate GTrack to the calorimeter and match them to MTC.
    • smt_hit, smt_d0hit, smt_hitalgs - Fix SMT cluster error bug.
    • gtr_htf - Fix processing of p12+ Monte Carlo.
  • Functionality Improvements
    • AATrack - Install v00-02-01.
    • cft_reco - Split unphysical clusters.
    • d0reco - Add AATrack to OBJECTS / LIBRARIES (but off by default).
    • em_evt - Make access methods consistent for both DST and TMB.
    • geometry_db - Align calorimeter in Z with tracker ("3 cm shift").
    • gtr_find - Add option to run AA via command line control.
    • muonid - Added 2 variables (region, octant) into the thumbnail.
    • muo_segmentlinkedlist, muo_trackreco - Decrease the number of fakes due to the algorithm down to 0.1%,; increase speed; increase efficiency for low pT medium muons.
• p13.02.00 (Build 4)
  • Bug fixes
    • chpart_reco - Protect against divide by zero.
  • Functionality improvements
    • chpart_reco - Use get_AllPVertices and get_AllSVertices to get primary and secondary vertices.
    • vertex_evt - Support get_AllPVertices and get_AllSVertices for chpart_reco.
    • cps_reco - Set the threashold for single strips at 0.00125 MIPS/1000.
    • geometry_db - Install new muon MC geometries to support p12/p13 MC. New d0MC_v7_cftrot_muon_survey.evpack (p12.xx), d0MC_v8_cftrot_muon_survey_smtasbuilt.evpack (p13.xx).
    • muon_geometry - New MuoBaseGeometry_mcjun02_p12MM.rcp that contains p12/p13 Monte Carlo positions.
    • smt_lpos, smt_hitalgs - Make 90 degree errors for large clusters same as in p11.
• p13.02.00 (Build 5)
  • Bug fixes
    • cps_unpdata - Fix problems with CPS electronics map.
    • muonid - Fix truncation of variables Idnhit and Idndeck.
    • run_config_fwk, config_base, mag_field_config - Fix magnetic field for files with events from more than one run.
  • Functional improvements
    • bc_impTagreco - Implement proper track impact parameter significance resolution functions based on p11.09 data.
    • cps_geometry - Support new alignment.
    • caltables, calunpdata - New weights for massless gaps and middle hadronic with EM board readout.
    • caldata - Support for d0sim (new GainTable).
    • d0reco - Drop muon TrackChunk from TMB (no longer needed).
    • geometry_db - MC: Muon geometry for p12,p13, SMT geometry for p13, Calorimeter geometry for p12,p13, (same as data); Data: CFT geometry (Gaston Alignment), SMT geometry (centered with the CFT), Calorimeter geometry, CPS geometry.
• p13.03.00 (Build 1)
  • Bug fixes
    • bc_impTagreco - Protect against divide by zero in impTagAlg::calculate()
    • bc_impTagreco - Fix DCA significance calculation for pathological case. Cures infinite loop problem encountered will processing single particle Monte Carlo.
    • cps_reco - Protect against using invalid CPSCellId's, caused by corrupt CPS raw data. Using invalid cell id's resulted in RECO crashing in CPSReco.
    • cps_util
      • Permit access the database to get mips and gains for each channel if requested.
      • Assign default calibration constants to each channel, whether the calibration database is accessed. (Before the change, the default constants are assigned only if the database is not accessed.) I also c
      • Corrected the names of certain variables. (adc_per_mip --> adc_per_mev) 2. it
      • Deal with corrupt raw data. When unpacking an HDI channel, the code keeps an internal counter of SVX chip ID. It increments whenever it meets one new chip. If the new chip ID does not equal to the expected chip ID, the new code will skip the rest of the data in this HDI channel.
    • cps_unpdata
      • Changed the ADC per mip value by a factor of 1/2, for runs after 167475.
      • Do not access calibration database whenever the event is from MC.
    • unpack_mrc - Replace asserts with error logger messages in MRCCtrlModule if bad data is detected.
    • muonid - Protect against overwritting fixed arrays. Problem caused reco to crash on events with corrupt raw data.
  • Functionality improvements
    • muonid - Added/fixed thumbnail variables: tightMuotrack, segIndex, calnLayer, tanlam, partID.
• p13.03.00 (Build 2)
  • calorimeter_geometry / calroad - Fixes a bug in CellLine and CellPlanes which incorrectly assigned NORTH and SOUTH. Does not affect normal CalReco, but does affect extrapolation using CellLine.
  • cps_unpdata - Turn off debug printout.
  • geometry_db - New muon Monte Carlo geometry to support p13.
  • muon_geometry - New muon Monte Carlo geometry to support p13.
• p13.04.00 (Build 1)
  • Bug fixes
    • bc_impTagreco
      • Fix memory leak in impTagTrackCategory.
      • Fix inconsistency in tag contents (mis-match between track indices on one hand and corresponding impact parameters, errors on the other).
    • em_util - Stop building shared libraries.
    • muo_caltrackreco - Fix missing quotation mark in MuoCalTrackReco.rcp causing "email danielw@fnal" error message.
    • muonid - Changes for thumbnail: add region and octant, pack tanlam with mantissa, change scale of xA,yA,zA, increase nbr of bits for nhit,ndeck,sctimeA, sctimeBC, qptloc Change default value of ChargedParticleIndex from -1 to 255.
    • vertex_analyze - Increase quark block to prevent overwrites.
  • Functionality improvements
    • cal_nlc - Use caladdress directly for calorimeter addressing.
    • cft_reco - Revert back to not splitting unphysical CFT clusters.
    • config_base - Add two muon quality words to Global object.
    • d0reco - Include CalNadaChunk in TMB+.
    • geometry_db - Include improved SMT and CFT alignment (both internal and SMT to CFT).
    • gtr_find, gtrprop, smt_path - Turn on energy loss. Improves low pt pulls.
    • fps_reco - Fix clustering for real data.
    • fps_unpdata - Fix electronics map.
    • jetreco - Add nItems to thumbnail.
    • mag_field, mag_field_config - Remove inadvertent 1cm shift in solenoid Z position.
    • muo_analyze - Add two variables that are in the thumbnail.
    • muo_hitreco - Add two muon quality words to Global object. This will allow the muon group to investigate the frequency and crate location of out-of-sync problems.
    • muo_trackreco
      • If 2 local muons share hits they are considered as coming from the same physical muon; one needs criteria to chose the BEST. Until now the criteria was using (among other things) the pT of the local muon track. Detailed studies show that this introduces an important bias in pT. This will be changed to use the chi2 instead of the pT.
      • When trying to make a muon track, if the track length exceeds ->5 meters (rather than 30m!) then stop the fit.
    • recocert - Add tracking efficiency histograms.
    • semreco - Takes the shift calo/tracker from calorimeter_geometry instead of applying a hard-coded shift if the event is found to be real data.
    • vertexutil - Fix error in the way phi is calculated for certain tracks (wrap around bug). It has the effect of needlessly eliminating a number of tracks from vertex inclusion. Fixing it increases the number of found vertices 10-20%.
• p13.04.00 (Build 2)
  • Bug fixes
    • cps_evt - Replace throwing exception with error logger message.
    • fps_unpdata - Fix installing fps_map_v04.txt in rundata area.
    • fps_unpdata, fps_reco - Fix dynamic range so FPS clusters can appear in the thumbnail.
    • jetreco - Fix bug with processing the same event two times in a row.
    • thumbnail - Fix memory leaks.
  • Functionality improvements
    • l1l2trigbits, event_info, l1l2_evt, l1l2_reco, d0reco - Include L1/L2 trigger information in the thumbnail.
• p13.04.00 (Build 3)
  • Bug fixes
    • l1l2_evt, l1l2_reco - Fix bug with unpacking L1/L2 information.
    • muo_centralmatch - Fix dealing with failure to match when going from muon system to central system (try central to muon).
  • Functionality improvements
    • gtr_analyze - Change the beam offset to (0,0).
• p13.04.00 (Build 4)
  • Bug fixes
    • muonid - Protect against overwriting fixed arrays.
    • thumbnail - Fix bug affecting CPS clusters in the thumbnail.
  • Functionality improvements
    • d0reco - Include CPSClusterChunk in the DST.
• p13.05.00
  • Bug fixes
    • cal_nlc - Fix so that coef file is picked up in current directory; change code to send error message and throw exception if coef file not found. Fixes problem with calorimeter non-linear corrections in p13.04.00.
    • gtrbase - Fix memory leak in GTrack::get_num_measurements() - does not affect farm production but does affect default user production.
    • gtr_find - Fix a bug in the handling of smt 1d clusters so that refit AA tracks can be saved correctly.
    • muonid - Fix memory leak when unpacking MuonParticle from thumbnail; fix bug that screwed up the Chp index of the muon into the tmb in some cases; fix backward compatibility to deal with muon charge in thumbnail.
  • Functionality improvements
    • l1l2_evt - Per request from Top group, the number of L1Cal towers was increased from 15 to 45.
• p13.06.00
  • Bug fixes
    • cft_calibration - Bugs were exposed when a new version of d0stream with a new memory management was released. The bugs were fixed in this version.
    • em_evt - Remove _pshwrE from EmTmbObj to prevent crashing in maxopt / IRIX; take the EMparticle direction from the best match track; tighten the match between the SEMparticle candidate and a scone cluster by requiring both hottest calo. tower to be close; remove message "Ou est donc le cluster ????"; return a 0-size vector for empty road (take care of negative E).
    • em_util - Fixes for gcc build.
• p13.06.01
  • Bug fixes
    • cal_nlc - Fix bug related to processing Monte Carlo events. At a very small rate, energy in the calorimeter would be greatly increased due to a bug with inverting the non-linear corrections. A detailed description of the problem is available.
    • d0reco - Include support for processing MC and perserving information about the MC hits.
    • emreco
      • EMTrackFinder: take the EMparticle direction from the best match track (correct a harmless bug since in most cases only one track is concerned).
      • EMClusFinder : tighten the match between the SEMparticle candidate and a scone cluster by requiring both hottest calo. tower to be close. Bad matches were are the origin of the message "Ou est donc le cluster ????"
    • gtr_htf - Fix rare floating point exception crash when processing Monte Carlo events.
    • thumbnail - Fix bug in ThumbNailL3Util causing RECO to crash when processing events taken with p13 Level 3.
• p13.07.00
  • Functionality improvements
    • The following improvements allow RECO to access the calibration databases
      • calibration_management - Disable CPS access to db until db is restored.
      • cft_calibration - New CFT db interface for improved speed.
      • cft_unpdata - New CFT db interface for improved speed; remove global ADC cut (20) when using db.
      • d0reco - Support calib db access for SAM jobs.
    • recocert - Added SMT histograms; improve tracking effic histograms; bug fixes.
  • Bug fixes
    • bc_eTagreco - Protect against very rarely occurring problems when unpacking TMB files (crash when trying to use an invalid LinkPtr).
    • l1l2_evt, l1l2_reco - Fix L1/L2 information in thumbnail for Monte Carlo.
• p13.08.00
  • No RECO changes.

User Information

• Currently available RECO rcp's do not support reprocessing Monte Carlo trigsim output so that l1l2chunk is available in the thumbnail. (Jodi Wittlin).

  When reprocessing trigsim output with reco, the l1l2chunk is not produced in the thumbnail. This can be gotten around by adding to the *_reprocess.rcp the line: RCP unpack = <d0reco reco_unpack_mcraw>.

• Duplicate muon local tracks (Christophe Clement)

  In p13 you can/should use the track group variable: tr_group. The muons are grouped into groups of muons sharing 2 hits or more. In the MuoTrfit block there is now the 'tr_group' variable. This variable is the group index. For each group, there is only one muon with tr_group>0 which is the best muon in the group all other muons will have a group index of -(tr_group). So to summarise: to remove double counting you should only take local muons with tr_group>0.

• CFT axial-only tracks.

  Starting in p13.01.00, CFT axial-only tracks that were not connected with either CFT stereo hits or SMT hits are saved in the GTrack 401 chunk, and thus become ChargedParticle's (implying they are available in the thumbnail). These tracks are not used by vertexing algorithms, but are available for user analysis. The tracks may be distinguished from 3D tracks as follows:

  • GTracks - axial-only tracks have z = 0 and tan(lambda) = 0 at DCA.
  • ChargedParticles - axial-only tracks have z = 0 and tan(lambda) = 1000.

• Muon geometry

  The muon geometry for data processing does not contain the best alignment constants for runs taken after the July, 2002 shutdown. "This is not considered to be a disaster, since the differences should be small. So far we do not have quantitative results on the size of the differences, nor on the effect of such differences on the local muon reconstruction and matching. Hopefully negligible." -- Martijn Mulders. This will be rectified in p14.

This page maintained by Harry Melanson