240 likes | 401 Views
ECAL Monitoring (non-hardware). Outline. ECAL monitoring introduction Data taking for ECAL monitoring Monitoring data processing and data analysis Offline stability correction. Laser Transparency Correction. APD/PN_REF. Physics Event Time. Correction Factor. max. few %.
E N D
Outline • ECAL monitoring introduction • Data taking for ECAL monitoring • Monitoring data processing and data analysis • Offline stability correction ECAL Monitoring Review, Ocotber 2009
Laser Transparency Correction APD/PN_REF Physics Event Time Correction Factor max. few % Interpolation x Scale laser transparency change by constant “Alpha” to match tall response change Transparency Measurements 20 min Xtal response change under irradiation • The crystal transparency – and hence ECAL response - changes under irradiation. • The changes are on the order of few percent, they are transient on a time scale of hours. • Measuring and correcting these changes obviously implies that one measures the stability in general – any change in response will reflect in the laser transparency measurements. • The scale factor Alpha only applies to response changes due to transparency change. ECAL Monitoring Review, Ocotber 2009
The Gap in the LHC Bunch Train Orbit Gap WarningTestEnable TestEnable L1A GT issues these triggers. Abort gaps occur at ~10 kHz, 3564 bunch crossings, 756 empty bunch crossing, 118 empty bunch crossings at the end of the orbit Laser pulses at ~100 Hz • Use ~1% of gaps for the transparency measurement • LED stability pulsing (aka soak light) will use (a fraction of) the remaining gaps ECAL Monitoring Review, Ocotber 2009
Monitoring System Hardware 4 • Laser light sources see David’s presentations • Light distribution system (fibers, optical switches, diffusing spheres, etc.) • Very stable PN-diodes used as reference system (MEM) • Precision pulsing system for electronics calibration (separate hardware for MEMs) • LED pulsing system for the EE, injecting into level 1 fanout see Sasha’s presentation APD PN APD VPT ECAL Monitoring Review, Ocotber 2009
The ECAL Calibration Sequence • The monitoring is designed such that sets of 800/900 channels can be illuminated by the light sources. • All of ECAL is sub-devided into 90 such regions, 72 in EB and 9 each in EE. • Two EE regions are read out by two DCCs, resulting in 92 separate measurements to scan all of ECAL. • Currently we take 600 laser events to measure the transparency. • It takes a few seconds to switch the light distribution from one region to the next. • These few seconds are used to record test pulse and pedestal events. • We have four laser wavelengths and in EE in addition two LED wavelengths. • The composition of the sequence (number of LASER, TP, PED, wavelenghts) is configurable, allowing a wide range of options. • Depending on configuration a scan of the entire ECAL takes between ~15 (just blue laser and minimal switching time) and ~40 minutes (two wavelenghts, extended switching time). The optimal composition of the sequence is currently being discussed and will likely be optimized over the years. This optimization will likely only reflects on the per-mille level of the ECAL performance. ECAL Monitoring Review, Ocotber 2009
ECAL Laser Monitoring Online Data Flow at P5 Global Trigger Disk Buffer CMS – Mostly ECAL (also DT, TK, HCAL) LaserSystem TTCci EMTC ECAL ECAL/HCAL RCT/GCT Gap Events 100 Hz Physics Events 100 kHz Storage Managers LaserFarm ECAL Gap Data Readout Units Online PCs All Gap Data HLT Like Gap Filter Dropping FEDs Split / Reduplicate Filter Farm T0 ECAL Monitoring Review, Ocotber 2009
Calibration Sequence Operation & Responsibilities The data taking chain has been fully commissioned and debugged. It works. Future development is restricted to adding features or adjusting to changes in the cDAQ level and possibly some tweaking on the level of the DCC. Maintenance and operation is expected to be very limited and routine operation only. • ECAL DAQ level : configure sequence composition, configure (some) laser settings (mainly intensity), ensure proper operation of ECAL readout, Philippe Gras, ECAL DAQ experts • CMS DAQ level : distributing calibration trigger, gap real estate allocation, GT experts • Data handling at HLT level : HLT filter module maintenance, HLT experts • Data handling at the StorageManager level : handling streams, directing the streams to laser farm, mounting laser farm disks, DataOperations experts. • Laser Farm hardware (8 DELL PE2950, 32 cores, 20 TB disk) is under cDAQ responsibility. • Beyond P5, raw data handling by central DataOps, however for the time beeing data gets only stored, but not repacked and not reconstructed. Note : Getting to the current state was a MAJOR endeavor. Heroes made it possible ! Documentation : https://twiki.cern.ch/twiki/bin/view/CMS/EcalMonitoringControls ECAL Monitoring Review, Ocotber 2009
ECAL DAQ and The Sequence EMTC DAQ review : http://indico.cern.ch/conferenceDisplay.py?confId=63447 ECAL Monitoring Review, Ocotber 2009
EMTC – Interface ECAL DAQ to Light Sources Responsibility : Philippe Gras, Marc Dejardin (Saclay) http://indico.cern.ch/getFile.py/access?contribId=6&sessionId=1&resId=4&materialId=slides&confId=63447 ECAL Monitoring Review, Ocotber 2009
Data taking monitoring via DQM • Monitoring data from the calibration sequence is equivalent to physics data. • The ECAL DQM (online, quasi-online, private, offline) looks at monitoring data. • This allows to assess if proper monitoring data is taken and probes the functionality of ECAL in a quite detailed way. DQM code analysing the monitoring data is maintained by the ECAL DQM experts. DQM plots are beeing looked at by ECAL shifters, ECAL DPG and PFG experts. ECAL Monitoring Review, Ocotber 2009
Monitoring Data Flow from Raw to Offline DAQ GT LASER CMS Laser Farm Raw APD/PN Corrected APD/PN Etc. Gap Events OMDS Corrected APD/PN P5 ORCON Offline Tier0 CAF Offline Reconstruction ORCOFF • This processing of the laser data and determination of the correction factors is sometimes referred to as the ‘laser data analysis’. This is misleading. • The procedure illustrated here is envisioned to be fully automatic. ECAL Monitoring Review, Ocotber 2009
Laser Farm Data Flow Saclay group, Julie Malcles et al Monitoring Laser results Calibration data streaming MusEcal Sorted data Sorting Primitives Light checker Laser plots Fastcheck OMDS • The data arrives on the laser farm as raw data streamer files. • It needs to be repacked - equivalent to merged into PD on TIER0. • We choose to write the output of that step separate files for each set of LASER, PED, TP, LED data per FED. • These files get reprocessed and the relevant information gets stored in a root file, called the LaserPrimitives. All this information is stored in OMDS. • In parallel, the LightChecker performs a quick analysis, extrcts some book keeping information. This information is displayed on a web page. It is being currently integrated into the standard DQM. • The processing happens quasi-realtime. The delay between data taking and availability of the results at P5 is on the order of ~20 minutes. Its limited by the way the data is made available to the laser farm. ECAL Monitoring Review, Ocotber 2009
Laser Farm Operation & Responsibilities • The data processing is done with CMSSW jobs relying on the P5 releases and a number of scripts and deamons. • The operation is fully automatic. It works, some further optimization and reliability improvements need to be done. • The resource balancing on the laser farm needs needs to be optimized, most notably the fact that we have separate buffer disks on the PCs is non-optimal. Some discussions with cDAQ in the past how to improve. • All the code and scripts were written by the Saclay group, in particular Julie Malcles, Marc Déjardin, Federico Ferri, Philippe Gras, Gautier Hamel-de-Monchenault, Patrick Jarry. • All the jobs run under a user group ‘ecallaser’. In addition to the people above there are about a dozen people from the Saclay and Caltech groups member of the ‘ecallaser’ group. • LaserFarm by Saclay group on call service phased in during CRAFT09. Documentation : https://twiki.cern.ch/twiki/bin/view/CMS/EcalLaserMonitoring ECAL Monitoring Review, Ocotber 2009
Laser Farm Monitoring & Online Tools MusECAL November 21 2008, ~10 am • MusECAL : ROOT analysis package running on LaserPrimitives to extract history information, trends, systematic effects etc. Also contains a GUI to visualize all quantities which are stored in OMDS. Developed and maintained by the Saclay group. ECAL Monitoring Review, Ocotber 2009
Laser Farm Monitoring & Online Tools Light Checker • Light Checker : Displays status of data streaming and processing. Delay is ~minute for the streaming info. Note : There is several minutes delay of data being delivered to Laser Farm since T0 is served first. http://ecal-laser-room-03.cms/laser-fastcheck/ Developed and maintained by the Saclay group. ECAL Monitoring Review, Ocotber 2009
Porting of LightChecker to DQM The information displayed in the LightChecker is being ported to CMS DQM format by Massimiliano Chiorboli (Caltech). This will allow to store the info along with the rest of the DQM information. ECAL Monitoring Review, Ocotber 2009
Laser Challenge : Offline DB Tables OMDS : 1 Sequence, 92 transparency measurements, 92 timestamps to(1) t1(1) t2(1) ORCOFF : 1 DB ‘BLOB’, 1 timestamp, granularity = ECAL to(92) t1(92) t2(92) • Solution : Store three transparency measurements, typically ensuring that a linear interpolation between them actually covers the intervall of validity of the ORCOFF record. • Problem : o2o has to properly mapp the OMDS content onto the ORCOFF struture Not fully validated. Data in the past partly corrupted. • Real problem : We artificially create a finer granularity (LightMonitoring region) in ORCOFF than forseen by default (ECAL). • Additional temporary problem : Usage of non-unique time stamps (CPU time). ECAL Monitoring Review, Ocotber 2009
Actions and Responsibilities on DB • What remains to be done is the full validation of the transport of the corrections factors from OMDS to ORCOFF. • The code applying the actual offline correction is in place and works. It is rather simplistic. It likely is fully sufficient in its simplicity. • A significant fraction of ECAL people from the Monitoring, Offline and DB groups have worked on OMDS-o2o-ORCOFF chain in the past, in particular : Cavallari, Timciuc, Ferri, Gras, Orimoto, Hamel-de-Monchenault, Innocente, Gataullin, Paramatti, Veverka + many others + bunch of coordinators • A renewed effort now from the ECAL DPG Calibration conveeners in the context of the LaserChallenge (aka Laser end-to-end test). • It is the currently the only know reason why stability corrections can not be applied. ECAL Monitoring Review, Ocotber 2009
Second Wavelength Monitoring • The monitoring with blue light at 440 nm is sensitive to transparency change and almost all other response changes. • It is a known fact that other gain changes, in particular in the EE, will be equally important. The magnitude and time structure of effects varies greatly. • To correct these effects, monitoring with a second wavelength is needed. In EB this is provided by the infra-red laser – even though non-transparency related effects might be small. In the EE, non-transparency related effects are substantial, options are the orange LED and the green laser. VPTs are insensitive to red and IR laser • Details have to be studied in detail. This is among the primary task of the LED/VPT stability group and an important task for the monitoring group. Almost all technical aspects, except the light source and the soak light, are equivalent to the ones described in this presentation. The current proposal is that offline, all corrections will be merged into one single correction. ECAL Monitoring Review, Ocotber 2009
Second Wavelength Monitoring Chris Rogan : http://indico.cern.ch/getFile.py/access?contribId=1&resId=0&materialId=slides&confId=23924 ECAL Monitoring Review, Ocotber 2009
Example Performance CRAFT09 See Julie Malcles : http://indico.cern.ch/conferenceDisplay.py?confId=68676 Laser History 0.3% Magnet History • Before correction: mean stab = 0.75‰ • After correction: mean stab = 0.56‰ This is Laser Analysis ! 0.1% TP History More results coming from Patick Jarry et al ECAL Monitoring Review, Ocotber 2009
Monitoring in the ECAL Organigram DPG G. Franzoni (Minnesota) + T. Tabarelli de Fatis (Milano) Field Technical Coordinator W. Funk (CERN) + K. Bell (RAL) Trigger/DAQ A. David (LIP) + P. Paganini (LLR) Run Coordinator N. Cartiglia (Torino) + A. Bornheim (Caltech) Upgrade P.Busson (LLR) Deputy: R. Hirosky (Virginia) Reconstruction software F. Ferri p0 (V. Litvine) EB-EE LV Minnesota Singovski Cooling CERN (Pons) Common Software LIP Musella CSC Vlassov Clustering & corrections Y. Maravin Single electron (P. Govoni) LASER/Monitoring Caltech/Saclay Bailleux/Gras LED System Virginia Ledovskoy DCC LIP Da Silva Non-event DQM ETH/Notre-Dame Thea/Kolb Z->ee (R. Paramatti) Calibration/Aligment M. Gataullin R. Paramatti Phi symmetry (S. Argiro’) DCS & ESS ETHZ Dissertori Di Calafiori Data Links Minnesota Singovski TCC LLR Paganini ES DAQ Bialas DQM G, Della-Ricca, E Di Marco Laser Monitoring (A. Bornheim/P.Jarry) DCU Torino Arcidiacono SRP Saclay Gras PFG T. Orimoto, C. Rovelli ES Sensors CERN Peisert Low level calibration (J. Fay) EB-EE HV Roma/RAL Cavallari, Bell Databases F. Cavallari, G. Organtini Geometry/Aligment (B. Heltsley) ES System CERN Barney Crystals ETHZ/CERN Nessi-Tedaldi/Auffray Simulation B. Heltsley Endcap Performance (D. Cockerill/A.Ledovskoy) • Laser Monitoring DPG task : P. Jarry, A. Bornheim, as representatives of the Saclay and Caltech Groups. Both groups dedicated a signifcant fraction of the M&O quota to this. • In the future, the ‘Laser Monitoring’ task under DPG will focus on ensuring proper operation of the monitoring system and on performing higher level systematic studies. ECAL Monitoring Review, Ocotber 2009
Summary • Quote from CMS RunCoordinator : «The ECAL monitoring has come to maturity». • The monitoring procedure has been implemented and is operational. The last missing piece is the validation fo the correction factor transport from OMDS to ORCOFF. • Stability has been monitord at the level of 0.1% during CRAFT. Operation of the monitroing system is likely to expose many low level systematic effects over the years. • The available manpower during the commissioning and debugging of the system has often been stretched thin. • Given the complexity of the system, manpower to maintain it, ensure proper operation and continued optimal performance is substantial. Current main sources of manpower, the Saclay and Caltech groups, will ensure sufficient manpower is available. • For the future, main focus will be on higher level analysis of systematic effects. The manpower source for this may not be limited to the traditional players. ECAL Monitoring Review, Ocotber 2009