Calibration of CMS detectors CMS detector Calibration of

Calibration of CMS detectors CMS detector Calibration of each subdetector CMS global ‘calibrations’: B-Field Alignment Mumbai 24 Octoebr 2009 T. Camporesi, CERN 1

Exploded View of CMS Plus Side Mumbai 24 Octoebr 2009 Minus Side T. Camporesi, CERN 2

Calibration: why? • Every particle detector ‘signal’ is derived from phenomena which happens in the material of the detector following the passage of particle(s). In one way or another such interactions are converted into analog electrical signals first and then digitized for ease of handling by a computer code. • Calibration of a ‘detector signal’ amounts to understand what happens in the chain ‘effect in material’ > collection of electrical signal >conv. to digital>transfer to readout device so that we can extract the maximum of information from the ‘calibrated’ response of the detector. • The calibration parameters can be used directly at the time of the conversion of analog to digital or applied later in the software reconstruction of the event • In a second stage we want also to convert the measured signal into ‘physics observables’ units ( e. g Ge. V, # of electrons etc) Mumbai 24 Octoebr 2009 T. Camporesi, CERN 3

Calibration: how ? • Particles: collisions and/or beam backgrounds (e. g. halo muons) at LHC. Without to LHC beams, cosmics and/or test beams ( prior to installation) • Artificial signal sources: Lasers and LED pulses, injected charges ( AKA test pulses) Mumbai 24 Octoebr 2009 T. Camporesi, CERN 4

Constraint from LHC: timing • Bunch structure: beam bunches are separated by 25 ns. The readout of each detector of CMS is structured in ‘pipelines’ (sequences of data which are transferred from the front ends to the memory banks of the readout electronics) subdivided into ‘slots’ which are synchronous with the beam structure, ie. each contains info re a 25 ns period. • Typically a slot in a detector pipeline contains info re the digitized pulse height sampled in that 25 ns interval + some geographical info • It is clear that in order to properly associate the detector info to a given bunch crossing the ‘shaping’ of the analog pulses has to be fast in order to allow allocation to the right bunch crossing Mumbai 24 Octoebr 2009 T. Camporesi, CERN 5

Constraints from LHC: radiation • The radiation levels at LHC ( especially at high luminosity) can be very trying for any material. • While the choice of materials and readout component has taken into account the need to ‘survive’ for years to such radiation levels, appropriate monitoring capabilities had to be foreseen at conception to allow following response changes Mumbai 24 Octoebr 2009 T. Camporesi, CERN 6

Objectives from Physics • In other words how good should the calibration be ? E. g. – Higgs decay in gg: contributions to constant term of EM energy resolution <0. 5% – Muon momentum resolution <10% at P~1 Te. V/c (reconstruction of mass of Z’) translates into requirement on m-hit position resolution and chamber alignment – Efficiency at separating vertices close to beam line (pileup, heavy flavor identification): depends on tracker resolution and alignment Mumbai 24 Octoebr 2009 T. Camporesi, CERN 7

Whole CMS: reality since summer 2008 Mumbai 24 Octoebr 2009 T. Camporesi, CERN 8

Most calibration experience from CRAFT Mumbai 24 Octoebr 2009 T. Camporesi, CERN 9

SYNCHRONIZATION • The first fundamental requirement is to get the CMS triggers synchronized to the beam transit through CMS: in the process we need to take into account ‘time of flight’ (e. g HF is at 11 m from IP) and length of fiber carrying the trigger info. Once this is taken into account there is no substitute to real particles crossing the detector to validate (needs a ‘time scan’) • Then we need to get data pipelines synchronized with the trigger ( and hence amongst themselves): ability to reconstruct an event relies on the assurance that in a given slot (referred to the beam crossing structure) the information in each sub-dets pipeline refers to the same ‘interaction’. Here again TOF and fiber length are an ingredient to guess the right delay, but real particles are a must. • Delay (AKA latency) scans can be performed by each subsystem : typically a scan consists of ‘rough’ stepping in 25 ns jumps and then a finer adjustement in ~1 ns steps. • The metric used to define the synchronization is based on coincidences/time correlations: e. g. the delay of a trigger is defined by Mumbai 24 Octoebr 2009 T. Camporesi, CERN 10 looking at accumulations of triggers when defining t 0 by the beam

Synchronization example CSC triggers DT TPG BX DT triggers DT TPG BX CSC triggers After correction DT trigger phi DT TPG ~1 BX late wrt CSC bottom and ~2 BX late wrt. CSC top Mumbai 24 Octoebr 2009 T. Camporesi, CERN CSC trigger phi 11

PIXELs 3 barrel layers: 4. 3, 7. 2 , 11 cm 2 Ecap disks: R [6 -15]cm @ 34. 5, 46. 5 cm in Z • Pixel detector chosen to be readout on the n implant side to – Anticipate ‘type inversion’ due to non-ionising radiation effects ( after type inversion depleted region will be on ‘pixel’ side) – profit from large Ex. B effect on the drift of the electrons which provide the signal: charge from ionization will distribute over ~2 pixels and the sharing will allow improved localization precision (10 -15 mm vs >30 mm without charge sharing) Channels: 48 M+18 M • Calibration issues: – Lorentz angle – Charge calibration – Threshold setting to consider a pixel as ‘hit’ and transfer data to pipeline – Setting of gain and offset of front end amplifier+offsets of the various opto-transmitter/receivers along the data path to the readout module (analog data is transferred to the FED) Mumbai 24 Octoebr 2009 T. Camporesi, CERN – Noise measurement 12

Pixel Charge calibration • Essential to ‘equalize’ pixel response: a charged particle deposits ~22000 e-: this charge is deposited over a cluster of Pedestal=Intercept several pixels and the ( determined by Xtalk from readout resolution on the cluster position depends on the correctly measuring the charge sharing • Method: calibrated charge( 15% variation from pixel to pixel) injected at the front end amplifier. Will improve with real tracks Mumbai 24 Octoebr 2009 T. Camporesi, CERN Gain=Slope 1 VCAL = 65. 5 e- 13

Pixel readout Thresholds • The threshold is determined mainly by the Xtalk induced by the ‘readout’ ( clocking of data out of the front end buffers) which are mounted on the pixels M ea n = Th re sh Error function e=erf(VCAL) Mean noise: 141 e- Bpix 85 e- Fpix Si gm a = No ise old Mumbai 24 Octoebr 2009 T. Camporesi, CERN 14

Barrel Pixel: Lorentz angle • Ex. B effects cause drifting charge carriers to be deflected at an angle: this effect is beneficial for spatial resolution as it spread charges over several pixels, but requires a correction to be applied to the reconstructed cluster center. • The angle ( hence the correction) can be found by looking at tracks with varying incident angles and determine the angle where the cluster size is minimum Mumbai 24 Octoebr 2009 T. Camporesi, CERN B= 3. 8 T Endcap B= 0 T 15

Pixel: the end result Ddxy Mumbai 24 Octoebr 2009 Ddz T. Camporesi, CERN 16

Tracker • • 198 m 2 of silicon 9. 3 Million strip Classical p on n substrate implant. Readout via APV which are mounted on the detector modules: sample the signal and maintain it in an analog pipeline. Data are digitized by FEDs in two possible modes: Peak or deconvolution FED perform zero suppression, common mode subtraction Readout sampling Mumbai 24 Octoebr 2009 T. Camporesi, CERN 17

Tracker: pedestals and common mode • The baseline of the readout chip (ie. The average pedestal level for a set of 128 strips) is adjusted to be at roughly 1/3 of the dynamic range. The pedestal value for each strip is evaluated by taking a large number of randomly triggered (i. e. without any signal) events. The values determined this way are then used in the FED to perform Zero suppression. • The common mode noise is evaluated event by taking the median of the pedestal subtracted pulse height of the 128 strips and subtracted from the individual strip pulse height. • The noise depends on size of the sensors and on the broken wire bonds: the strips show distinct ‘footprints’ in this respect Those which show anomalous noise behaviour are masked 18 Mumbai 24 Octoebr 2009 T. Camporesi, CERN

Tracker: Zero suppression • The FED operates normally in Z/S mode (pedestal subtraction, common mode subtraction, identification of channel above threshold) • Threshold: 5 snoiseon single channels or 2 snoise on contiguous channels Mumbai 24 Octoebr 2009 Info being sent on the fiber from the front end to the FED T. Camporesi, CERN Used to synch channels 128 strips 19

Tracker: internal synchronization • Synchronization between front-ends is very important: signal amplitude is attenuated by 4% /ns when reading in deconv mode. • Synch is done using the tick marks at the end of the transmission frame for each of the tracker partitions Mumbai 24 Octoebr 2009 T. Camporesi, CERN TIB synch 20

Tracker: trigger synchronization • Two steps: – first synch in 25 ns steps with respect to external trigger – Then in 1 ns steps to find the best synchronization mean(14 ns) comparable to TOF from Muon chambers ( which triggered) Mumbai 24 Octoebr 2009 T. Camporesi, CERN 21

Tracker: transmission calibration • As analog information is transmitted optically, an essential part of the calibration consists in the tuning of the laser which drives the optical lines (transmission is affected by laser-to-fiber alignment, laser output spread, fiber to fiber alignment at the three patch panels ). This is done using the Tick mark reference ( square pulse assumed to be the same for all channels and equal to the max output of the APV 25 chip) to equalize the transmission of all the fibers. Byproduct of this is a calibration scale of 274 e - /ADC count. Mumbai 24 Octoebr 2009 T. Camporesi, CERN 22

Tracker: the result e=99. 8% Excluding known faulty channels Abs calibration 262+/- 3 e-/ADC count Mumbai 24 Octoebr 2009 T. Camporesi, CERN 23

Tracker : Track Finding Efficiency Tag and Probe method • Tag : Stand alone muons �dz�< 30 cm, �dxy�< 30 cm, �eta�< 1, 0. 5 < �phi�< 2. 5 (at point of closest approach) • Probe : Tracker reconstructed muons Combinatorial Track Finder (collision algorithm with special outside-in seeding) Cosmic Track Finder dedicated algorithm STA muon tag Efficiency (%) CRAFT Results for Approval CRAFT 09 CTF 99. 8± 0. 1 Cosmic. TF 99. 8± 0. 1 24

ECAL • PWO crystals: high density (8. 3 g/cm 3), small moliere radius (2 cm), ~26 X 0 in EB, ~25 X 0 in EE • Rad resistance: – Radiation hard: high light yield even after 10 years of LHC exposure – Radiation soft: rapidly looses transparency ( up to 20%) when exposed to high intensity rad flux. Requires constant transparency monitor Mumbai 24 Octoebr 2009 61200 Xtals in EB 2 x 7324 Xtals in EE T. Camporesi, CERN 25

ECAL: inter-crystal calibration • Like in any calorimeter the first and most fundamental calibration step is to equalize the response of each channel. The raw response of a given crystal (including also the analog readout chain) can differ channel to channel by up to 30%. • Only way is to use radiation signals: – Radioactive sources (~Me. V) at crystal QA bench: 4 -5% -All crystals – Cosmic rays(~250 Me. V) : 1. 5 -2. 5% All Barrel SM – e beam 120 Ge. V: << 1% 9 barrel SM + 160 EE crystals • Abs energy scale : 38 Me. V/ADC in EB Mumbai 24 Octoebr 2009 T. Camporesi, CERN 26

ECAL cosmic calib Arrangment of SM for cosmic calib Lower statistics (and possibly some fine containment effect) Mumbai 24 Octoebr 2009 T. Camporesi, CERN 27

ECAL: transparency monitor • Radiation affects crystal transparency not crystal light yield: recovery time relatively fast. Need continuous monitoring of transparency • Laser distributed to each crystal (split into groups of 200 fibers where laser intensity is monitored by means of a PN photodiode) • Laser Data taken during run (use of abort gap) to collect 100 Hz of laser triggers. Data payload is such that for every laser trigger one can read max ½ SM. Collect >500 laser event/crystals every ~30 minutes to provide crystal by crystal transparency correction Mumbai 24 Octoebr 2009 T. Camporesi, CERN 28

ECAL : laser calib OK A typical crystal RMS dist. For all crystals in EB Normalized (to PN diode which monitors Laser light) response of typical channel Mumbai 24 Octoebr 2009 T. Camporesi, CERN 29

ECAL: calibration success Expected 68% prob interval CRAFT 08 Validation of precalibration with avg m signal : rms 1. 1% Stopping Power Minimal interval containing 68% of distribution (asymmetry in energy deposit distribution due to lack of containment of showers likely for high P m) Radiation processes Collision processes Mumbai 24 Octoebr 2009 T. Camporesi, CERN 30

HCAL Mumbai 24 Octoebr 2009 T. Camporesi, CERN 31

HCAL : calibration • Hcal readout: Charge to Digital conversion done at front end. QIE is a 5 bit ADC switched into 4 ranges ( covering a dynamic range of ~10000) • First requirement: calibration of charge to ADC counts. Done on each chip by injecting calibrated charges for the 4 ranges. This gives the f. C/ADC ratio for each channel. • Second calibration of ADC to Ge. V: all HCAL scintillator are crossed by sources driven inside thin tubes which are mounted at installation time. The cross calibration of source deposits with real particles has been done for a sample of source exposures on modules exposed to the testbeam. This calibration has been extended to the totality of the detector through the source exposure. • The ADC to Ge. V correction could be cross checked with the Cosmics exposure in 2008 ( and also using the Beam splash events in september 2008): this comparison ( possible only on 80% of modules, due to the directional flux of cosmics) showed that the original calibration was good to 10% and with the cosmics could be reduced to 4% Mumbai 24 Octoebr 2009 T. Camporesi, CERN 32

HCAL: cosmics calibrations Mumbai 24 Octoebr 2009 T. Camporesi, CERN 33

HF • Quartz fibers embedded in a iron matrix. Showers provoke Cerenkov light in the fiber. Light detected by PMTs • High radiation resistance, low light yield: need > 3 Ge. V of deposited energy (e- beam) to get 1 Photoelectron Mumbai 24 Octoebr 2009 T. Camporesi, CERN 34

HF calibrations • • HF is calibrated both with Rad Sources (driven along the Quartz fibers) and with LED pulses flashing on the PMTs LED allows Single PE measurements, hence not sensitive to Quantum efficiencies of PMT windows light collection efficiencies. The correlation between the two methods is good for most channels. Discrepancies have been confirmed by beam splashes, hence calibration used is the one form the sources. Mumbai 24 Octoebr 2009 T. Camporesi, CERN 35

Muons: DT • Drift tubes: convert drift time to space. Drift ‘trajectory’ depends on B field • Vdrift depends on pressure, temperature, and gas composition Superlayer: 2 r-f chambers and 1 z chambers Mumbai 24 Octoebr 2009 T. Camporesi, CERN 36

DT calibration • Key to achieve physics goals (<10% momentum resolution at ~1 Te. V is alignment –see later) • Determination of ‘T 0 wire’ ( corresponding to time of ionization deposited close to wire) is one of the most important parameter used in the conversion of Drift-time to space • Drift velocity is calculated from size of cell/max drift time • Vdrift monitored by dedicated small chambers located at gas mixer output Mumbai 24 Octoebr 2009 “time box” t 0 Total drift time Track position=Time x Driftvelocity Time= TDCTime - t 0(wire) - t. Trig (SL) - tprop @ CRAFT: Different values top-bottom due to TOF @ LHC: Flat distribution vs Sector number T. Camporesi, CERN #Sector 37

DT and B field in yoke gaps is almost negligible with the exception of Inner Layer in external wheels ( MB 1 in YB+/-2) Angle of track wrt to perpendicular Effect of 0. 5 T // to wires Mumbai 24 Octoebr 2009 T. Camporesi, CERN 38

Muons: RPC CMS choice: avalanche mode • Signal due to ionization is collected within ~10 ns: hence very fast detector, ideal for triggering, with relatively low spatial resolution. • Sensitivity to pressure, temperature: Mumbai 24 Octoebr 2009 T. Camporesi, CERN 39

RPC Calibration • Monitor noise levels (mask noisy strips!) • Signal large enough to allow common threshold throughout system • Scan Voltages and thresholds and monitor efficiency (extrapolation from DT tracks) Mumbai 24 Octoebr 2009 T. Camporesi, CERN Red Th. 210 m. V Green Th. 220 m. V 40

RPC Gas Monitoring • It is essential to keep the quality of the gas mixture (96. 2% C 2 H 2 F 4, 3. 5% i. C 4 H 10 and 0. 3% SF 6 )under control. The mixture is recirculated through purifiers filters with an injection of 8% fresh gas • To do so the gas mix is checked by – gas chromatography, PH and flouride monitoring) – A set of three RPC chambers in a cosmic ray telescope , sitting on the surface and fed respectively with fresh gas, gas form the recirculation system before and after the purifiers filters Mumbai 24 Octoebr 2009 T. Camporesi, CERN 41

Muons: CSC Y determined by position of wires group (wires are ganged in groups of 12) hit X position by reconstructing Mumbai 24 Octoebr 2009 T. Camporesi, CERN centroid over three strips 42

CSC: strip response equalization • The baseline ( pedestal) is estimated by taking the average pulse height of the two samples preceding the pulse (rms between 2 and 4 ADC counts) is subtracted from the Pulse height measured in the 3 subsequent samples • A precision pulse ( 1%) is injected at the front end to estimate the channel to channel gain variations and the ratio wrt to the average gain is used to weight the charge measured by a given strip. Mumbai 24 Octoebr 2009 T. Camporesi, CERN Gain distribution very gaussian 43

CSC: Xtalk • Cross talk between Strips is an important effect in CSCs • Correction well under control : Xtalk matrix involving signals in central and side strips applied on the raw charge collected in 3 time bins in the 3 strips used to evaluate the centroid position Mumbai 24 Octoebr 2009 T. Camporesi, CERN 44

CSC: Lorentz angle compensation built in Mumbai 24 Octoebr 2009 T. Camporesi, CERN 45

CSC: close to design performance • CRAFT 08 Mumbai 24 Octoebr 2009 T. Camporesi, CERN 46

Global ‘calibration’: B field inside solenoid • B field mapped by specially designed ‘mapper’precisely inside solenoid prior to insertion of detectors ( Tracker, ECAL, HCAL): precision << 10 -3 verified with cosmic tracks in CRAFT 08 Mumbai 24 Octoebr 20 T. Camporesi, CERN 47

B field in Iron yoke • Field in Iron slabs of return Yoke are modeled using a finite element code TOSCA and assuming azimuthal simmetry. Model estimation can be checked ( sector by sector) using cosmic tracks – Comparison of tracker extrapolation to impact point on first layer of DT (MB 1) sensitive to field value inside first yoke layer – Comparison of angular deviation between straight segments (field in open regions where DT are inserted is negligible) reconstructed in successive DT stations (e. g. Mb 1 vs MB 2 sensitive to field in second yoke layer and so on) • This checks revealed large discrepancies ( up to 30%) between model and ‘observables’ • Traced to feature of TOSCA in dealing with boundary condition Mumbai 24 Octoebr 2009 T. Camporesi, CERN 48

B field in Yoke Layers Craft 08 Data Wheels • After properly defining the boundary condition the map agrees with the cosmic ray tracks measurements within 8 percent in the worst case. (residual % difference blue *) • Scaling factors determined per wheel, sector and layer using cosmics are ‘calibrating’ out the residual discrepancies (red *) • The CRAFT 08 data highlighted the need of detailed modeling of sectors where ‘phi’ symmetry assumption was obviously bad (highlighted by green boxes) Mumbai 24 Octoebr 2009 T. Camporesi, CERN 49

Global ‘calibration’: alignment • Alignment of tracking detector components is fundamental to maintain best muon P resolution: need to know absolute position of tracking elements with a precision better than the given detector resolution • Difficulty due to bulk of material which is interspersed between tracking elements: for ex. in the muon chamber case multiple scattering effects can cause ‘distorsions’ of tracks which are several cm while one aims to locate elements with precision of order 100 mm Mumbai 24 Octoebr 2009 T. Camporesi, CERN 50

What needs to be aligned Mumbai 24 Octoebr 2009 T. Camporesi, CERN 51

Alignment strategy • CMS has built into the detector assembly mechanical structures which allow monitoring, through light beams, of ‘large’ movements due to B-field, thermal expansion, mechanical settling and so on • Ultimately the final alignment is based on using tracks (so far cosmic tracks). Approach based on – Align locally ( meaning over region of space where material effects are ‘manageable’) sub-detector elements using tracks, possibly with constraints coming from optical/mechanical measurements done at construction/assembly time – Align with respect to global alignment frame ( defined by tracker) locally aligned ‘elements’ treating them as rigid body Mumbai 24 Octoebr 2009 T. Camporesi, CERN 52

Hardware alignment Ä3 Independent Internal Alignments: Tracker, Muon Barrel, Muon Endcap ÄLink system to relate TK & Muon System Link Disk Alignment Rings Lateral half view B A R R E L 3 r-z alignment planes with 600 staggering in f. MAB 12 carbon-fiber bars (z-bars) attached to the vacuum tank Rigid carbon fiber structures “Modules for the Alignment of the Barrel” (MAB) attached to the Barrel wheels LEDs at DT chambers & Z-bars observed by cameras in MABs LINK System Transparent Amorphous Silicon Position Detectors connected by laser lines Rigid reference structures Alignment Rings , Link Disks & MABs Complemented proximity Mumbai 24 Octoebrby 2009 electrolytic tilt-meters, T. Camporesi, CERN sensors, magnetic probes and 53 temperature sensors

Methods: Align. parameters Track residuals • Minimize: • Millipede: Takes into account align parameters and track parameters simultaneously (ends up solving a matrix of dimension ~105 using various ‘tricks’ and approx) • Hit & Impact Point (HIP): Iterates ignoring correlations linked to track parameters aligning each module and refitting tracks after each iteration (at each iteration amounts to solving a matrix of dimension 6 for the 6 alignment parameters of each module) • CRAFT allowed understanding that best approach is to use the two in cascade profiting from the complementary strengths of Mumbai 24 Octoebr T. Camporesi, CERN 54 the two 2009 approaches

Alignment: Tracker& pixel • Use ditribution of Median of residuals (DMR) after alignment as indicator of alignment precision (the median is insensitive to multiple scattering and depends less on non gaussian tails than the mean) CRAFT 08 • RMS of DMR gives ‘quality’ of alignment Example: Pixel (reinstalled in shutdown) did indeed move between ‘ 08 and ‘ 09 CRAFT 09 data CRAFT 08 alignment aligned DMR RMS (μm) Data 2009 Alignment prelim MC alignment Ideal alignment Modules >30 Hits BPIX (x) 2. 6 2. 5 2. 1 757/768 BPIX (y) 4. 0 2. 5 2. 4 757/768 FPIX (x) 13. 1 13 12. 0 9. 4 391/672 FPIX (y) 13. 9 13 11. 6 9. 3 391/672 TIB (x) 2. 5 3 1. 2 1. 1 2623/2724 TOB (x) 2. 6 3 1. 4 1. 1 5129/5208 TID (x) 3. 3 4 2. 4 1. 6 807/816 TEC (x) 7. 4 8 4. 6 2. 5 6318/6400 Alignment quality Confirmed in 2009 Mumbai 24 Octoebr 2009 T. Camporesi, CERN 55

Alignment: Muons, DT • Using a Millipede algorithm Means of residuals of DT segments Relative Z of superlyers: tracks vs photogrammetry after before Mumbai 24 Octoebr 2009 T. Camporesi, CERN 56

Alignment: muons, CSC Mumbai 24 Octoebr 2009 T. Camporesi, CERN 57

Alignment of CMS Alignment of DT wrt to tracker Mumbai 24 Octoebr 2009 T. Camporesi, CERN 58

What happens with beam ? • For tracker and pixel : collisions will provide tracks which will quickly allow cross checking/improving cosmics calibrations and alignment ( methods substantially the same ) with added value of well known and stable phase of particle timing with respect to digitizers clocks • For Calorimeters: – Use specific particles: p 0 for ECAL and isolated tracks for HCAL – Use azimuthal symmetry of summs over min bias events to equalize response in ‘detector rings’ – When more lumi use known ‘candles’ : Z e+e- • Data flow catering to calibration needs: special ALCA streams dedicated to fast processing meant to feedback Al &Ca constants with minimal latency ( 24 Mumbai 2009 to reco jobs T. Camporesi, CERN 59 4824 Octoebr hours)

Not covered because of lack of time • Jet energy scale: calorimetric measurement affected by fluctuations due to jet particle composition, poor match between e/h response of ECAL and HCAL, eta dependance of material in front of calorimeters and of material and space (in 4 T field) between the calorimters. • Continous developments to try to improve on this limitations trying to use tracking and energy deposition patterns information • … but this is easily the matter for another looong talk … Mumbai 24 Octoebr 2009 T. Camporesi, CERN 60

Summary • CMS is one of the most sophisticated particle detector ever built, despite its ‘functional’ simplicity • The care with which it has been built and the sophisticated monitoring and software is allowing to fulfill. . . and in certain cases exceed its design goals as far as physics potential is concerned. • Calibration and alignment are the first step towards excellence and , despite having reached already impressive goals will continue to deserve priority still for some years to come Mumbai 24 Octoebr 2009 T. Camporesi, CERN 61

alignment Are you sure your code got the right minima

Backup slides Mumbai 24 Octoebr 2009 T. Camporesi, CERN 63

Barrel shrinking due to B-field - 2009 Movements of the Z -bar LEDs vs #run in the Z direction YB-2 YB+2 15° relaxation effect @ 75° beginnig of 0 T (~8 -10 h) 75° Final closure of Barrel @ 2 T 135° Final closure 3. 8 T reproducibility: ~< 100 microns 195° switching-on effect!!! max 200 microns O: 1 day long 255° 1. 8 T 0 T 3. 8 T 315° 0 T 1. 8 T 0 T 315° 0 T 3. 8 T 0 T 0 T