Online and Data Quality Monitoring Eduardo Rodrigues University

Online and Data Quality Monitoring Eduardo Rodrigues University of Glasgow LHCb VELO Software meeting, CERN, 6 April 2009 Online coordination: Kurt Rinnert Data Quality coordination: Eduardo Rodrigues Contributors: Eduardo (cluster monitoring), Sadia (track monitoring), Marco & Silvia (vertex monitoring), Marco (alignment monitoring), Malcolm (monitoring occupancy in HLT), Sara (Noisy & Dead strips), Abdi (noise vs voltage), James & Abdi (Landau fitting), Grant (Gain monitoring), Tomasz (TELL 1 algorithm monitoring), Paula. Eduardo Rodrigues

Monitoring: outline q A lot of recent work and progress – thanks everyone ! q Monitoring algorithms: - Most algorithms fairly stable - A couple of new kinds on the block: e. g. gain monit. , “expert” noise monit. q Monitoring analysis scripts: - area with main effort a lot of new tools available q Wiki pages with documentation and How. To’s are also being updated Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

But we’re not quite at this stage … Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Monitoring: algorithms packages “VELO moni group” Velo. Rec. Monitors v v Package for “high-level” (= ZS) data 6 monitoring algorithms in CVS - 3 related to cluster monitoring: “standard”, “expert”, for HLT - 1 for time-alignment studies (sampling) - 1 for track-based monitoring - 1 relevant to monitoring of closing v v Alignment monitoring part of track-based monit. algorithm Vertex-based monitoring taken from Tr/Track. Monitors Velo. Data. Monitor v v v Package for NZS data Several algorithms for error banks / noise / pedestals / TELL 1 s / … monitoring New algorithm for gain monitoring Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Monitoring: scripts package Vetra. Scripts “VELO moni group” q Centrally collect scripts, macros, Python modules for monitoring and data quality analyses q Area on critical path a couple of months ago q Glad to see so much effort went into this recently: - 37 macros/scripts in CVS! (we may even need some clean-up due to large overlaps in code) q If you still have handy private scripts then let us know … ; -) q And please commit your code, even if not in its “final” version (remember it’s a CVS repository) q http: //isscvs. cern. ch/cgi-bin/cvsweb. cgi/Velo/Vetra. Scripts/? cvsroot=lhcb Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Online presenter (1/2) q Monitors ZS as well as NZS data q 2 monitoring tasks exist: - Velo. DAQMon: clusters, noise, TAE, errors - Velo. Beam. Mon: beam position for closing - track-based & alignment monitoring will either be added to Velo. DAQMon or will get its own task q VELO online monitoring depends on Vetra project - has to follow faster release cycle of Online project by patching the current installation : S Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009 Kurt

Online presenter (2/2) q Now much easier to efficiently declare several pages (using new handy tools from Nicola for manipulating Hist. DB) q Available pages at present: default cluster monit. , TAE monit. and per TELL 1 noise pages q Implementation of pages is not yet complete. But it’s much easier compared to the past q Page layouts and histograms display options will evolve once we have user feedback Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Online presenter: snapshots (1/2) Cluster monitoring Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Online presenter: snapshots (2/2) Noise monitoring Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Online presenter: cluster-based monitoring Offline Presenter Yes q # of strips per cluster Yes q Cluster ADC value (all sensors) Yes q Cluster ADC value (R sensors) Yes q Cluster ADC value (f sensors) Yes q # of clusters for sensor “i” Yes No q # of strips per cluster versus sensor Yes q Cluster ADC value versus sensor Yes q Active chip links versus sensor Yes q # of clusters per event 1 page in presenter (1 plot produced per sensor if n. Clus>0) Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Online presenter: track-based monitoring Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Online presenter: track-based monitoring 5. Module mismatch Presenter layout 6. Biased residuals vs sensor 7. Pseudo-efficiency 6 5 8 7 8. Local f coordinate pop-up pages when clicking on each histo: 5. Original histo + “Used sensors” histo 6. The 8 histos related with residuals mean and s for R and f and A- and C-side 7. (clicking disabled) 8. Original histo + eta & f histos Eduardo Rodrigues (global coords. ) LHCb VELO Software Meeting, CERN, 6 April 2009

Online presenter: vertex monitoring 1. PV left-right delta x Presenter layout 2. PV position x 3. PV position y 1 4 2 3 4. PV position z q Track. Vertex. Monitor in Tr/Track. Monitors q (Need to include this algo. In the Vetra monitoring sequence) Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

New Velo. Expert. Cluster. Monitor algorithm q New algo in Velo/Velo. Rec. Monitors for “expert” cluster monitoring q Creates histograms of the ADCs for each sensor q Makes plots for raw and on-track clusters q Splits sensors into different regions of r, q and f q Histograms are viewed by new disp. Cluster. py script (in Velo/Vetra. Scripts) Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009 Abdi, James

New disp. Cluster. py script (1/2) q Displays plots of the ADCs of clusters and the number of strips in a cluster q ADCs are fitted with a Landau-Gaussian convolution to find MPV and FWHM Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009 Abdi, James

New disp. Cluster. py script (2/2) q Also plots MPV vs sensor number q ADC values are angle corrected in the case of on-track clusters Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Gain calibration (1/3) Monitor the gain of links: Initially through by looking at link headers Next will look at test-pulse data Grant, Kazu Can the “full header swing” (FHS) be used for this purpose? As long as low/high header distributions are distinct (like plot below) this is a clearly-defined and stable quantity. Goal is to calibrate links: Can we somehow normalize link gains to some standard value? Would be useful for eventual d. E/dx measurements Eduardo Rodrigues Full Header Swing (FHS) “Header-Low” distribution LHCb VELO Software Meeting, CERN, 6 April 2009 “Header-High” distribution

Gain calibration (2/3) Code to make these plots works with Vetra v 7 r 0, and is in CVS of Velo/Velo. Data. Monitor, with ROOT macros in Velo/Vetra. Scripts. Can be used to quickly spot bad links ROOT code in Velo/Vetra. Scripts is able to plot: FHS, FHS scaled by link noise Average header value, average header value scaled by noise RMS of HH, HL distributions Link headers with unusually high (>600) and low (<400) ADC counts Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Gain calibration (3/3) Unusually high/low ADC counts: Avg headers: Link header distributions: Eduardo Rodrigues RMS of HH/HL distributions: LHCb VELO Software Meeting, CERN, 6 April 2009

VELO monitoring & analysis GUI Eduardo, Karol, Kurt, Paula q Consensus on how to run monitoring/analysis jobs: 1) run a PVSS recipe 2) “on button” run of a corresponding Vetra task 3) GUI to ease the display/analysis of histograms produced q Status of matters 1 month ago: - PVSS recipes exist since long (renamed recently) q In the meantime: - PVSS button to run Vetra task defined by Karol - New VELO monitoring GUI - has received many contributions ever since Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Velo. Moni. GUI: snapshots (1/2) Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Velo. Moni. GUI: snapshots (2/2) Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Next on the TODO list (1/2) q Update on Vetra/Brunel monitoring sequences: - Not all algorithms set to run in our monitoring sequences - On-/off-line specific options to be checked (keep in mind that one needs. opts files online!) q Installation of Vetra and monitoring packages for shifters: - Regularly updated - But needs better handling: e. g. , never change anything locally that is not in CVS or immediately committed to CVS ; -) - Some issues for the online version? q Discussion of shifters tasks: - we need to start thinking about our monitoring procedures at the pit - important for VELO shifters … and you’ve just learned this morning that we will have shifts very soon to start the training Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Next on the TODO list (2/2) q We need to collect reference plots! - If you haven’t done so yet, please get in touch with me to get this sorted by the end of this week q Tutorials: - We recently had (on the 13 th March) a first Vetra tutorial - More tutorials are foreseen, with special emphasis on monitoring q Documentation – my usual blabla: - Update twiki, remove old & obsolete pages, converge on overlapping pages, etc. - There’s been some activity. To be continued … Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009

Outlook (apologies if I’m not aware of very latest updates) Algorithms in CVS Analysis script(s) Included in GUI Doc in code/twiki Reference Tested in Vetra plots Responsible person Cluster monitoring Eduardo Track Monitoring Sadia Vertex monitoring Marco, Silvia Alignment monitoring Marco HLT occupancy Malcolm Noisy/dead strips Sara Noise monitoring Abdi, James Gain monitoring Grant TELL 1 Eduardo Rodrigues monitoring Tomasz LHCb VELO Software Meeting, CERN, 6 April 2009

Any comments? Eduardo Rodrigues LHCb VELO Software Meeting, CERN, 6 April 2009