Calibration chain developed at LAPP for the ECAL of CMS

The experience of large calorimeters, in passed and actual experiments shows that the calibration is an extremely long and methodic task, implying several redundant methods in hardware and software as well. If, as expected nowadays, the Higgs has a mass under 150 GeV, its discovery in γγ decay by LHC experiments will rely heavily on the resolution of the ECAL. The precise measurement of these two gammas will depend in particular upon an accurate calibration (the constant term is expected < 0.5% in CMS) and upon a precise γ/e energy scale. The knowledge and the mastering of the calibration will play a major role. We describe here the electronic system, that has been developed at LAPP to update permanently the calibration of the electronic readout chain of CMS-ECAL, during data taking. 1-The CMS ECAL and its read out The target characteristics of CMS ECAL make this project very challenging in terms of resolution, hadron rejection, particles isolation etc...Its characteristics are described in several publications [1,2,3] and I will only remember the principles of its read-out . As this crystal calorimeter is fully immerged in the magnetic field of 4 Teslas, the choice was made to read out the crystals by APD photodetectors. A very aggressive R&D has been sent since 4 years and is being concluded by the production of 120 000 of these detectors. The first pre-production is actually under tests. Each crystal will be equipped with 2 Hamamatsu APD’s, in parallel, and the read out is made by sets of 2 read out cards, grouping 5 channels each. On the read out card are implemented : A/ 5 chips “FPPA”, developed in Harris UHF1X process, which are multi-gain shaping amplifiers (gains = 1, 4, 8, 32) with a specific logic to indicate the first amplifier that does not saturates for the sampling (40 MHz) being made. The chosen amplifier is marked by 2 bits, added B/ 5 commercial flash ADC 12 bits (Analog Device AD 9042, previously developed for USAF, in Bipolar XFCB Process), that digitize the amplitude every 25 ns. C/ 5 serializers that convert the parallel 20 bits words (40 MHz) to a serial stream out (0.8GB.s) for the opto-coupler. D/ One opto-coupler, developed in CHFET ASGA process, that transmits the digital information through a set of 3 fibers to the Read Out System ECAL (ROSE) card that, in the control room, links the front electronic to the general DAQ and to the trigger. E/ One “Control Chip” (CTRL) developed in DMILL BICMOS process, that receives, from the ROSE card, through optical fibers, the adjustment parameters for the readout. It deserializes and dispatch them to individual chips. Apart from the DAC, all the chips are full custom and have been developed at CERN by Princeton group. The final version of the whole system has been put in test beam for the first time in August and the analysis is under way. 2The calibrations of ECAL The task can be divided into 3 general topics [4]: 1. Before being mounted on the detector, all the supermodules will be pre-calibrated on a dedicated electron test beam to provide an initial precise set of calibration coefficients for each channel. 2 . During data taking, the day to day, channel to channel calibration inside a same region will be updated during sterile cycles of the machine by light injection and charge injection. This calibration is made by alignment of the response of individual channels belonging to a same region. A region can be defined as a group of channels sharing the same bunch of fibers. 3 . The medium and long range (in time) absolute calibration of the whole detector will rely on W and Z decay measurements. The calibration of ECAL is then organized on several partially redundant methods: 1. Injections at the inner edge of the crystals of light pulses produced by two lasers and passing through a fiber distribution system: This system will follow the response of the ensemble crystal-readout chain to light pulses along the time. The first pulse, at the middle of the scintillation spectrum (420 nm) will follow the shifts of the overall system (in particular due to loss of transparency of crystals under irradiation)[5,6]. The second one, at 600nm is almost insensitive to transparency losses and will follow the APD gain and electronic shifts. 2 . An independent system of charge injection at the input of the read-out chain will follow the shifts of the readout chain and cross-calibrate the different gains of the preamplifiers. 3. “In situ” calibration with physics events ; at low energy, the isolated electrons will be measured precisely by the tracker and their momentum compared with the response of the calorimeter. At medium energy (around 20 GeV), the W’s and Z’s will provide an absolute calibration for the whole detector. According to luminosity, this “in situ” calibration will need between a few days and a month during which the local calibrations will have to be adjusted. 3-The calibration by Charge injection The project consist of adding on the readout card a) one Test Pulse Logical System (TPLS), implemented inside the CTRL, b) one DAC, c) 5 injectors, producing on request a current pulse at the input of each FPPA. These pulses have a shape identical to the APD’s one and their common amplitude is proportional to the order given to the DAC. The system has been developed at LAPP (Annecy) in DMILL process, with following requirements : 1. no perturbation of the characteristics of the chain (in particular no add of correlated noise, no cross talk, nor any deterioration of characteristics). 2. at least as precise and reliable as the readout chain 3 . robustness to eventual shifts due to irradiation damages 4. Use of existing supplies (0/5V) and very low power consumption 3.1 The Injector The injector is the most critical part of this system. It has to cover the total dynamic of the APD’s (pick current of 4mA, and a full-scale charge of 48 pC) with a linearity better than 0.5 % ; It has to be radhard up to 10 n/cm and 3 Mrads in γ (10 years of LHC irradiation). The shape is a fast negative fall followed by an exponential rise (τ=15ns), with a technical dispersion in amplitude and time lower than 10%. The PSRR (Power Supply Rejection Ratio) against the supply and the DAC line has to be as low as possible. Since 1998, we have made 5 generations of prototypes ; a first one in AMS to test the principle, then two in CMOS, one in CMOS Bipolar and a last version integrating the 5 channels in one chip. In the final design, the injector is made of 3 components : 1. The trigger part, that adapt the PECL window signal (400ns width) received from the CTRL to levels compatible with the pulse generator stage. The edge of signals are also slightly reshaped to avoid the time jitter in the trigger of calibration pulses. 2 . The amplifier that translates the voltage level produced by the DAC into a current level, transmitted to the pulse generator stage 3. The pulse generator is mainly a two branch bipolar circuit. The trigger splits the flowing current from one branch to the other one and the capacity Cout (60pF), previously charged is rapidly discharged and then recharged through Rout (250 Ω) and the preamplifier (Zi ~7 Ω). The decay time of the pulse is completely fixed by Rout * Cout . For precision and flexibility, the Cout has been placed outside the chip. Figure 1 Principle of the Injector 3.2 The DAC In a previous version, we have tested a commercial circuit, Analog Device AD 8582A in CB CMOS technology. This DAC is a parallel input, 12 bits, 0/4Volts, 5mA output, supplied on 0/5Volts, its dissipation is 5mW at rest. Its reference voltage level can be measured on an output pad. This chip has been extensively tested in lab and in irradiation beams. It fulfills the calibration requirements . In a second step, a 10 bits DMILL version has been founded in view of assembling the whole project in one chip. This DMILL DAC is still under investigations due to process difficulties during foundry. 3.3 The control chip The control chip first amplifies the amplitude of the 40MHz clock and of the data’s received from the optocoupler; then it recognizes, interprets and dispatches the orders to the different parts of the readout card. 4-The R&D 4.1 The tools Figure 2 shows the chain that has been mounted at LAPP to measure the linearity, the dispersion and the stability of different components of the project. This chain is based on a PC linked to a VME and CAMAC crates for charge measurements. It is also linked to a Keithley Multimeter 2020 by GPIB link for levels measurements. An overall monitor Labview® program makes systematic rampings on the DAC with charge and levels measurements at each step. The linearity of the injector has been measured in tension (at the edges of Rout), which is the most precise and in total integrated charge, after Cout In some runs, we have also verified by on a LECROY 9361 linked by GPIB that the shape was not affected. Figure 2 Principle of the Measurement chain For irradiation tests, we moved an identical chain on the different available sites. Figures 3 : set up used for irradiation tests After a preliminary test on the Nuclear plan Ulysses at Saclay (IR1), we have used the SARA installation at Grenoble (France) (IR 2) and after it’s shut down, the CERI installations at Orleans (IR 3,4 and 5), where the irradiation facility had been moved. These beams [7,8] produce neutrons by stripping of deutons on a thick beryllium target. The average energy of these neutrons is ≈ 6 MeV and the FWHM is ≈ 6 MeV. Theses beam have a photon contamination that has been evaluated to 3.6 kGy for an integrated dose of 2*10 n/cm . The fluence depends on the distance to the target and can be adjusted up to 10 n/cm/ hour. We have also used the 72 MeV proton beam of PSI (Villigen, Switzerland) (IR 6 and 7). The correction parameter for protons is ≈ 2 and the equivalent integrated dose for this irradiation was 4 10 n/cm2. We have also used a photon Cobalt source facility at PSI to