VELO Software Update TELL 1 Algorithms Bitperfectness Beetle
VELO Software Update • TELL 1 Algorithms Bit-perfectness Beetle header correction • Error analysis • Occupancy/Noise analysis • Gain Calibration • Parameter Storage • TELL 1 clustering in simulation • Pile-up Kurt Rinnert (for Chris Parkes)
TELL 1 Algorithms Tomasz, Guido + Karol, Kurt, Alex… • In June TED run significant discrepancy seen between emulation and TELL 1 – TELL 1 producing significant extra ‘noise’ clusters • TELL 1 C lib and firmware 4. 0 recompiled and installed – Test data with various algorithm configurations & databases taken. – Major problem solved – Some discrepancies still under investigation on boundary strips in FPGA processing, working with Alex Gong – Currently 96. 7% bit perfect 2
Beetle Header Cross-talk Tomasz Szumlak • Cross-talk in first channel of each data link from header bit on Beetle chip • Correction algorithm in TELL 1 Cross-talk In data • Significant for ACDC 3, Small effect in final system now correct timing & link compensation – Prob. turn-off correction for 2009, but monitor ACDC 3 Data Before correction After correction Channel from same link Draft LHCb note circulated for comments 3
Error Monitoring Chiara Farrinelli • TELL 1 Error monitoring developed – Main purpose - identifies synchronisation errors – Beetle pipeline column number decoded based on headers and compared with reference Beetle on TELL 1 • Plots added to standard monitoring • Tuning procedure for header high/low states and hgh rate tests planned 4
Occupancy & Noisy Strips James Keaveney (+ Sara, Kurt, CP) • Procedure to monitor & mask out any noisy strips or links • 1. Occupancy – Clustering algorithm strip inclusion thresholds are set based on noise in strips – Occupancy monitoring developed – Strips appearing in clusters above an occupancy threshold (currently 1%) in noise data can be masked out in TELL 1 • 2. Noisy/ Dead Strip identification – Strips with higher noise identified – Dead strips identified using noise data with no voltage applied to detector – this makes them more obvious – monitoring of both classes put in place 5
Gain Calibration Grant Mc. Gregor • Before • Gains in velo system calibrated using Beetle headers • Cross-checked with testpulses • Now looking at d. E/dx 6
Gain Calibration Grant Mc. Gregor • After • Gains in velo system calibrated using Beetle headers • Cross-checked with testpulses • Now looking at d. E/dx 7
Parameter Database (with ST) • Identified need for – Storage of calibration parameters used in data taking • e. g. pedestals, thresholds, timing parameters… • Plan to automatically store at end of run in XML file – Storage of parameters for uploading/monitoring • Calculated parameters from analyses (pedestals…) that will be uploaded • Parameters needed for system monitoring (e. g. dead, noisy strips) 8
TELL 1 Clustering in Boole Tomasz Szumlak • Implementation of final VELO clustering in Boole • Old clustering • TELL 1 Clustering • 99. 9% works, timing OK • BUT problem with 4 -strip clusters (in TELL 1? ) causes crashes in associators – Was aiming for release last month, dealyed due to TELL 1 bitperfectness issue 9
Pile-up Serena Ogeero working on pile-up software Velo-ers happy to support when can • Decoding added for digital data Serena • Material description of pile-up carbon Mark & Tom fibre supports • Work started on including pile-up analogue data in Vetra emulation Tomasz 10
Summary • Other topics not covered here – Vetra v 7 r 2 released (Tomasz) – Monitoring Improvements (Eduardo, Kurt) – Closing improvements (Malcolm, Sophie, Marco) – Delta-rays (first work from James, David) – Preparation of CCE analysis, PVSS panel for data taking (Stefano) • Critical path for data readiness – Bit-perfectness – Complete error monitoring – Databases • Critical Item on Simulation – TELL 1 Clustering in Boole • Getting ready for TED run next week Thanks to Kurt for giving this, apologies as have teaching - Chris 11
- Slides: 11