The construction and commissioning of the CMS Silicon Strip Tracker
334 th International Conference on High Energy Physics, Philadelphia, 2008
The construction and commissioning of the CMS Silicon Strip Tracker
Giacomo Sguazzoni (for the CMS Silicon Strip Tracker Collaboration)
INFN Sezione di Firenze, via Sansone, 1 - I-50019 Sesto F.no (FI) - ITALY
As the start up date for LHC approaches, the detectors are readying for data taking. Here a review will be given onthe construction phase with insights into the various difficulties encountered during the process. An overview will alsobe given of the commissioning strategy and results obtained so far.The CMS tracker is the largest silicon microstrip detector ever built. Consisting of three main subsystems, InnerBarrel and Disks, Outer Barrel and End Caps, it is 5 . . with more than 15000 detector modules.The various integration procedures and quality checks implemented are briefly reviewed. Finally an overview is givenof checkout procedures performed at CERN, after the final underground installation of the detector.
1. CONSTRUCTION AND INSTALLATION
The CMS Silicon Strip Tracker (SST) is the world’s largest Silicon strip detector with a volume of approximately23 m instrumented by 15 148 modules for a total of 198 m of Silicon active area and 9 316 352 channels with fulloptical analog readout [1, 2]. The SST, shown in Figure 1, covers the radial range between 20 cm and 110 cm aroundthe LHC interaction point. The barrel region ( | z | <
110 cm) is split into a Tracker Inner Barrel (TIB) made offour detector layers, and a Tracker Outer Barrel (TOB) made of six detector layers. The TIB is complemented bythree Tracker Inner Disks per side (TID). The forward and backward regions (120 cm < | z | <
280 cm) are coveredby nine Tracker End-Cap (TEC) disks per side. In some of the layers and in the innermost rings, special stereomodules (Fig. 1, right panel) are used to build double sided assemblies able to provide three-dimensional positionmeasurement of the charged particle hits.The SST is made up of relatively small carbon fiber assemblies supporting several modules to ease the handling andthe mounting. These substructures are fully equipped with cooling and readout electronics to allow for stand-alonetests to be performed. This modularity has been the key-factor for the successful assembly of the huge number ofSi-strip modules [3] and all other ancillary components into the final detector. The following subdetector-dependentsubstructures can be found in the SST (Fig. 2): the TIB is split into 16 half cylinder shells each holding 135 to 216modules; the TID is structured in 3 rings per disk, each designed to support 40 or 48 modules; the TOB is madeof 688 rods , drawer-like structures providing support for 6 or 12 modules; and the TECs are made up of 144 petals per endcap, each holding 17 to 28 modules. During each stage of the assembly, every single component is verified.
TIBInner Barrel
TIDInner Disks
TEC Endcap
Tracker Support TubeTOB Outer Barrel
L~5.4m ! ~2.4m mm mm z view TIB TECTOB r ( mm ) ! z (mm) ! TID (a) (b)
Figure 1: (a) Sketch of the tracker layout (1/4 of the r-z view); (b) a TOB module on its aluminumtransport plate.
The CMS silicon strip tracker is the largest device of its type ever built. It is divided into four mainsubsystems: the Inner Barrel (TIB), the Outer Barrel (TOB), the Inner Disks (TID) and the Endcaps (TEC)(Figure 1). There are 24244 single-sided micro-strip sensors covering an active area of . Through-out the tracker, the strip pitch varies from the inner to the outer layers (from 80 µ m to 205 µ m ) in order toaccommodate the anticipated occupancy and to ensure a resonably uniform R − " resolution [1].The size of the device has led to a design where the basic unit, called a module, houses the siliconsensors and the readout electronics. Signals are collected via a CR-RC shaper, sampled and stored in ananalog pipeline by the APV25 front-end chip [2]. The APV25 chip also contains a deconvolution circuitto reduce the signal width [3]. The read-out can be performed either with or without the deconvolution,depending on the pile-up conditions. The two modes of operation are respectively called deconvolutionmode and peak mode.One aspect of the commissioning of the CMS silicon tracker will be the absolute synchronization ofeach module from data, to accommodate both the delays introduced by the hardware configurationand the effects due to the time of flight of particles. The objective is to be optimally synchronized withthe bunch crossing to maximize the efficiency while minimizing the number of fake hits from adjacentbunch crossings. This aspect is critical due to the high frequency of interactions at the LHC (nearly40MHz) and the width of the signal pulse ( > ).The CMS tracker is not able to produce a trigger signal by its own. An external (Level-1) trigger gener-ated from the information collected by other subdetectors, mainly calorimeters and muon chambers, isfed by dedicated optical control links from the front-end controllers to the APV25 chips. Upon receptionof a trigger signal, data of the corresponding bunch crossing are read from the pipeline and sent to thefront-end drivers (FED) via analog optical links. This is where the analog-to-digital conversion is done.To achieve the synchronization of the electronics a dedicated programmable delay is available in thePhase-Locked Loop (PLL) that is embedded in the front-end electronic of each module. It allows to shiftthe clock and trigger signals by steps of 1.04 ns. The global latency with respect to the central triggeris compensated by the “latency” parameter of each APV25 chip. That parameter defines an offset inthe APV25 analog pipeline by steps of 25ns. Finally, the differences in length of the analog lines arecompensated by programmable delays at the input of the FEDs. FED delays are set according to thefiber length information that will be optically measured from the front-end to the FED after the finaldetector cabling and stored in the construction database.With the level-1 trigger rate 400 times lower than the bunch crossing rate, the front-end chip may notoutput data for a considerable time. The chip therefore outputs a synchronization pulse called “tickmark” every 35 clock cycles when there is no data to read out. The tick mark can be used to firstsynchronize all modules with each other to compensate for the length of optical and electrical controllinks, who propagate the clock and the trigger to the front-end, as well as the electronics latency. This isdone adjusting the PLL delays such that, at the end of the procedure, the trigger arrival is synchronouson all tracker modules and, because of the delay settings in the FEDs, also the analog signal arrival timeis synchronous on all FED channels. The module PLL and FED delay settings determined in this way2 Figure 1: The CMS SST: simplified view (left panel); a quadrant of the rz section (right panel, bold lines represent doublesided module assemblies). a r X i v : . [ phy s i c s . i n s - d e t ] O c t th International Conference on High Energy Physics, Philadelphia, 2008 ~70cm ~1.1m ~80cm~1.1m
Figure 2: From left to right: a shell of the TIB; the innermost ring of a TID disk; a rod of the TOB; a petal of the TEC.
Optical links are checked by measuring the output level of an auxiliary signal issued by the readout chips and modulesare checked using pedestal and noise data at full bias (400 V) to spot high voltage issues and bad channels [4, 5].Once the substructures were completed, system-wide tests were performed for which cooling was provided as in finalCMS operating conditions. Temperature probes mounted on the modules are effective for identifying cooling circuitproblems [6]. In many cases, a cosmic ray setup was implemented to measure the S/N ratio for MIPs and to performtrack reconstruction exercises, so to verify the overall mechanical precision at the sub-millimeter level.During 2006 all substructures were assembled to create the main subsystems in regional integration centers. Finallythe tracker subsystems were installed in the Tracker Support Tube at the CERN Tracker Integration Facility (TIF)between fall ’06 and March ’07. The tracker quality was excellent. The total fraction of bad channels was found tobe 0.21%: 0.07% due to module failures, 0.05% due to bad optical links and 0.09% of bad isolated channels, mostlypreexisting problems. The assembly procedures and the stringent quality control tests have been proved to be soundas the integration process caused almost no new defects.Once the assembly of the SST was complete, it was possible to evaluate the material budget by introducing intothe detector simulation all the last-minute details added during the construction. The total thickness as a functionof η reaches a maximum of about 1.8 X around 1 . < ∼ | η | < ∼ . channels per m : TIB 2.2, TID 1.1,TOB 0.52, TEC 0.35) yielding a very large number of service connections in a very small space. This resulted in thepreviously discussed material budget issue and in difficulties to integrate the services in the limited room allocated.In December ’07 the SST was installed underground in CMS and the full cabling and piping was completed in Figure 3: Material budget rz map as seen by straight tracks originating at the interaction point: blue tone represents averageof total upstream x/X (i.e. local density of photon conversion tracks); gray tones (shown only on the z > /X (i.e. local normalized density of photon conversion vertexes). The white dashed line indicates the regionwhere TIB/TID services are deployed. th International Conference on High Energy Physics, Philadelphia, 2008
Figure 4: A reconstructed cosmic muon track passing through the SST collected during the Global Cosmic Run.
March ’08. Unfortunately two serious hardware failures of the cooling system serving the tracker (November ’07 andMay ’08) prevented the checkout activities from being performed as scheduled, i.e., immediately after the installation.The entire cooling system needed to be refurbished and was available in June ’08. In the 2008 LHC run, the SST willbe conservatively operated at about +12 ◦ C instead of at the design temperature of − ◦ C; in any case, the latter isonly required at design luminosity. During the 2008/2009 LHC shutdown the cooling system will be commissionedfor cold operations.
2. COMMISSIONING
Immediately after the finalization of the assembly at the TIF, an extensive testing program know as the “SliceTest” was carried out. It consisted of cooling, powering and reading out approximately 15% of the entire SST. Ascintillator-based trigger permitted 5 million cosmic muon events to be collected at a rate of about 5Hz, filteringout low momentum muons by means of a 5 cm-thick lead shield placed on the bottom scintillator. The Slice Testallowed the first experience to be gained on commissioning, safety, controls, monitoring, data management, trackreconstruction and alignment [7].After the installation within CMS and once the cooling system was available, the SST joined CMS operations. Acosmic global run without magnetic field took place from 7 to 14 July ’08. The test was very important for CMS: theSST readout makes up approximately 70% of the entire DAQ system. Despite the tight schedule for the first roundof commissioning, the SST performance is remarkable. The SST was smoothly read out within the global CMS DAQand several million events were collected. 79% of the SST modules were switched on (technical issues prevents theendcap on the − z side from being used) and, of those, around 93% passed the testing procedures first time. Thevast majority of the small number of modules excluded from this first global run will be simple to recover. In fact,the following subsequent more detailed commissioning performed before the closure of CMS in fall ’08, certified thatthe fraction of bad modules in the entire SST is below 1%.Cosmic muon tracks have been observed in the global run: an example is shown in Figure 4.
3. CALIBRATION HIGHLIGHTS
The optimization of the SST depends on several calibration steps [7]. The most relevant are briefly described here.34 th International Conference on High Energy Physics, Philadelphia, 2008 de l a y on P LL s [ n s ] module position in the control ring -latency [ns] a m p li t ude [ A DC c oun t s ] Figure 5: The delays set on the modules programmable delay units against the module position within the control ring (leftpanel); CR-RC pulse shape reconstructed by scanning on the latency values (right panel).
The SST dimensions and cable/fiber lengths imply non negligible timing differences compared to the LHC bunch-crossing period and the detector shaping time. Each module needs to be synchronized to compensate for pathdifferences of control signals (and for any electronics delay) as shown in Figure 5, left panel, where the delays seton the programmable delay unit of each module is plotted against the number of upstream controller chips. In fact,to limit the number of optical links, the control circuit is implemented by daisy-chaining several controller chipsintroducing a delay that needs to be compensated for.The analogue signal optimization is another fundamental calibration step, to be obtained by setting up the workingpoint parameters of the front-end chips and the downstream optical chain. Among these, one of the most importantis the configuration of the front-end chips with the appropriate latency value that, once a L1 trigger is issued, makeseach module send the data samples corresponding to the correct bunch crossing to the FEDs (the SST ADC boards).The result of the corresponding procedure ( latency scan ) are shown in Figure 5, right panel: the CR-RC pulse shapeis reconstructed by scanning a range of possible latency values in steps of 25 ns and taking as the optimal setting theclosest value to the maximum amplitude. In addition a fine tuning within a single bunch crossing will compensatefor time-of-flight effects [8].
4. CONCLUSIONS
The SST has been built, installed and read out within CMS. Commissioning procedures, online and offline taskshave been demonstrated to work well. Final commissioning and checkout procedures have demonstrated that the SSTquality is excellent with more than 99% of good modules. The SST is ready to deliver optimally reconstructed tracks.
References [1] S. Chatrchyan et al. [CMS Collaboration], JINST 3 S08004 (2008).[2] CMS Collaboration, CERN-LHCC/98-6 (1998); CMS Collaboration, CERN-LHCC/2000-16 (2000).[3] M. Krammer [CMS Tracker Collaboration], Nucl. Instrum. Meth. A (2007) 766.[4] C. Civinini [CMS Tracker Collaboration], Nucl. Instrum. Meth. A (2007) 726.[5] S. Mersi [CMS Tracker Collaboration], Nucl. Phys. Proc. Suppl. (2008) 318.[6] F. Palla [CMS Tracker Collaboration], Nucl. Instrum. Meth. A (2007) 104.[7] C. Delaere [CMS Collaboration], CERN-CMS-CR-2008-002.[8] C. Delaere et al.et al.