1 Status of preprocessing of data from TB DESY Tadeáš Bilka, Peter Kodyš, Peter Kvasnička Charles University in Prague F2F tracking workshop, May 30-31, 2016, Wien Charles University Prague

2 Outline 1.Data – resources for analysis (Peter Kodys) 2.First sweep and statistics (Peter Kodys) 3.Alignment (Tadeas Bilka) 4.Problems (Peter Kodys) 5.Plan

3 Data – resources for analysis People who helped directly with this task: Hua, Ulf, Hao, 2xPeter and Daniel 1.Data moved and collected on the ipnp30 Prague server, the other store should be a grid site in DESY 2.Anyone from the collaboration can ask for an account there, contact person: 3.17 external users and 3 from Prague have access now source “sroot.gz” files, 2438 converted “root” files 5.Available sets of files: Source files “.sroot.gz” (complete and done) Converted files to root “.root” (complete and done) New converted, sure full set “.root” (complete and done) SVD Global hardware “.xml” files (complete and done) Eudet telescope files “.raw” (complete and done) BonnDAQ files (in preparation by Benjamin, Florian and Botho) Quality check histograms and summaries (complete and done) 6.We are preparing table for usability for further analysis in basf2 for alignment, tracking, corrections,…

4 First sweep and statistics (preliminary) Full set of data: events in 4.17 days of clean acquisition time Biggest run 178: events Longest run 178: 14.9 hours Biggest subrun (collected in 1 file): events Longest subrun (collected in 1 file): 46.1 minutes Highest rate: 3000 events/sec For basf2 physics tests: 2 runs with secondary target lead 5mm in front of the telescopes kEvents in 17.6 minutes (rate 1200 events/sec) Run 339 unstable rate Magnet ON Beam energy 4 GeV Full PXD and SVD work Telescopes OFF (?) What is missing – tasks for the next TB: Magnet OFF Different beam energies Different angles Telescopes ON High statistics

5 First sweep and statistics (preliminary) Candidates for deep analysis: GeV, MagOff, no Pb, full PXD 565 kEvents 2.339, GeV, MagOn, full PXD, Pb kEvents GeV, MagOn, no Pb, full PXD, 599 kEvents GeV, MagOn, no Pb, full PXD, kEvents GeV, MagOn, no Pb, full PXD, kEvents GeV, MagOn, no Pb, high rate (2kHz) kEvents GeV, MagOn, no Pb very good statistics kEvents GeV, MagOff, no Pb extreme good statistics kEvents – 3GeV, MagOff, no Pb 6247 kEvents Benjamin request (97) - 5GeV, MagOff, no Pb work with telescopes (only one seems) 539 (447) kEvents

6 First sweep and statistics Example of summary table, many runs without logbook info: No info of run Only runs with secondary target Trigger rate problem

7 First sweep and statistics Example of summary table, selection for full runs:

8 Problem! Mass analysis cannot be proceed, because: 1.After clusterizer step following “VXDTF” produce errors and segmentation violation! 2.Another module “GenFitter” also crushed and produce segmentation violation! Please authors clean and correct it!

9 First sweep and statistics Example of quality check: Missing PXD- SVD correlation Correlations for PXD- SVD with secondary target in U Trigger PXD-SVD problem Good PXD-SVD correlation plot Good trigger rate in correlation region Rate of triggers in time – some effect from loss of synchronization?

10 Changes in VXD reconstruction for the testbeam: General, PXD General 1.Testbeam-contributed code in basf2 svn – All testbeam-specific code is in the testbeam package. – Modifications made to pxd/svd packages are good for Belle II – Some clearly marked testbeam-related code clones in pxd (unpacker) – We skipped the use of the git repository - not enough determination... 2.Changes to make reconstruction robust to malformed data – To make HLT/ExpReco processing stable even with experimental configurations, the PXD/SVD digit sorters now handle (some sorts of) malformed data. PXD 1.PXD simulation changed according to Benjamin Schwenker’s proposal – See the talk in the Monday PXD session – Modified to better reflect physics and DAQ configuration parameters – Adapted recently to better fit actual PXD DAQ settings and testbeam data 2.A lot of work in the unpacker - due to problems with DAQ – Decoding tables for strips, shifts in event numbering etc. – Special testbeam version of the unpacker

11 Changes in VXD reconstruction for the testbeam: SVD SVD 1.Introduced “three-consecutive- samples” filter – To pass the filter, the 6 APV samples have to contain 3 consecutive samples above threshold. – The filter effectively reduces noise digits – It suppresses off-time signals slightly more. 2.Import run-dependent hardware test data – Bad strips, strip noises, waveform shapes 3.Masking – List of hardware defects (GIacomo Caria) – Pre-run HW tests (Hao Yin, P. Kvasnicka) – Run-time masks (P. Kodys) – Need to find how to use in clustering

12 Geometry + Mapping → Tracking We directly imported Beast Phase II geometry with small modifications After SVD and PXD mapping was corrected and correlations seen, tracking started to work Visual check of residual distributions → PXD displaced by 1 mm w.r.t. SVD along Z→ absorbed into PXD sensor displacements Even with large misalignment and non-optimized sensors, (some) tracks were found and fitted. Beast geometry checked, only mapping needed hardware and software experts to work together. Good message for Beast Phase II 1 mm PXD z- residuals (unbiased, after Kalman fit) Plot from Track Fit DQM:

13 Initial Alignment & Issues Can be done with first messy data, but needs manual playing with Millepede (can get rid of fake tracks during iterating alignment and removing outliers) Many SVD sensors cannot be aligned due to low statistics, sometimes PXD not in data → big complications for alignment parametrization (e.g. which sensors to fix) – easy to introduce incosistencies in DB (alignment is defined relative to displaced geometry), fixed 1st and last SVD layer for alignment with no field Even after alignment, VXDTF gives plenty of non-sense to track fitters Working on some way to get rid of them for analysis + we should look into the VXDTF (sector maps problem?) Hits on GBL fitted tracks, p-value > ! Noise on slanted SVD added to tracks (you see hits extrapolated to the infinite slanted plane)

14 Preliminary Results PXD layer 1 residuals in run 340 (magnet + secondary target) – only high quality tracks, all cluster sizes, after initial alignment

15 Preliminary results Same run, same tracks, SVD layer 3

16 Alignment & Database We started with framework fall-back DB (nominal alignment) After run 96 (no magnet), DB with alignment was produced and used for fitting at HLT/expreco DB was distributed over . Only local DB used during whole TB After reprocessing we will provide a complete local DB at ipnp30, including e.g. masking Should go into PNNL central DB

17 Plans & Conclusions (Tadeas) Once full data quality analysis is finished, we will produce alignment data for all usable runs (and upload to PNNL)  Perfect exercise for calibration framework and anyone interested in PXD/SVD calibration, database, Beast Phase II, real data … Message to Beast Phase II and next beam test  Geometry fine, just check the mapping, track finding will work and we can do alignment  In next beam test, we need DB operational to test the calibration framework and distribute calibration constants to analyzers

18 TB DESY 2016 produced tracks with correlation in “off-line” check in basf2 delayed by hours to one day – still a bit long Source of delay: data transfer to a visible place on the net (ipnp30) On-line monitor was tuned and should be usable Very poor logbook, sometimes not editable for shifters because “open with other one”, only small subset of runs marked in, mistakes inside Missing in the logbook is a set of basic fixed-format information immediately available for parsing for analysis: Which subdetectors are working Status of subdetectors/planes Telescope IDs Magnet status Beam parameters Secondary target present Unclear naming convention: “physics”, “beam”, “test.all”, “test.pxd”, “test.svd” Communication between subgroups: geometry vs. pixel/strip counting and directions of axes at every step of data chain – needs to be solved before Belle II experiment From the point of view of basf2 Problems

19 Event mismatches from different sources, reasons partly unclear Workarounds on HLT, unpacker and special modules Mapping, coordinate flipping Remapping not existing in DHH firmware -> consequences for ROI on HLT and PXD Unpacking Extend ROIs on HLT -- hampering ROI selection Remapping during unpacking + coordinate flips Remapping triggered other errors in software (PXD Sorter, clusterizer) -> problems in express reco, no DQM ordering of DHE data/DHE-ID was inverted (increasing order expected as for ROIs) In unpacker not checked -> fixed meanwhile Minor issues with data format from DHH (not as defined in specs) -> Trigger Errors and Warning in PXD Unpacker fix firmware, change Warning back to Errors in software From the point of view of basf2 Problems

20 Thank you for your attention Plan (Peter) Continue screening the data for usability for analysis Cross-check basf2 results with Benjamin’s results Create tools for Belle II based on TB DESY 2016 experience Check usability of shape and eta corrections Tracking/alignment/database improvements Prepare tools for the next TB DESY end of 2016 Benjamin, Christoph, Carlos: coordination of TB analysis between groups, first regular meeting would be Friday 9h00. Carlos: preparing new Test beam in autumn of 2016