Università di Pisa. Facoltà di Scienze Matematiche, Fisiche e Naturali. Corso di Laurea in Fisica. Anno Accademico TESI DI LAUREA

Transcription

1 Università di Pisa Facoltà di Scienze Matematiche, Fisiche e Naturali Corso di Laurea in Fisica Anno Accademico TESI DI LAUREA TileCal la sezione centrale del calorimetro adronico di ATLAS: funzionamento e prestazioni CANDIDATA Deborah Capecchi RELATRICE Dott.ssa Chiara Roda

2

3 Introduction On the 4th of July 2012 the experiments ATLAS and CMS, installed along the LHC accelerator at the CERN laboratory, announced the discovery of a new particle capturing the attention of public and media. Despite the spread enthusiasm for the news, little has been said about the huge amount of designing and continuous renewing that lies behind this discovery: state of the art technologies, must mediate between scientists expectation and experimental difficulties and sometimes even unexpected situations. The object of this thesis is the new design of the low voltage power supplies which operate in the central part of the ATLAS hadronic calorimeter (TileCal), with attention to the testing and characterization phase of the production to which I participated personally. To understand the requirement on the design and the motivation for the redesign project, part of the work is dedicated to the understanding of the environment where the supplies work. The first chapter describes the main characteristic of the ATLAS experiment and of the collider on which the apparatus is installed. The Large Hadron Collider (LHC) is the most powerful particles accelerator ever built. It collides protons or lead ions at energy never reached before and it allows us to peer in a completely unexplored physics, opening the way to both predicted and unexpected results. ATLAS (A Toroidal LHC ApparatuS) is one of the two general purpose detectors installed inside the LHC tunnel and it is specifically designed to detect the presence of new unknown particles or phenomena generated during collisions. However LHC s power comes at a price, the excellencies of the accelerator, as high collision energy and event rate, arise at the same time experimental difficulties and set strict constraints on the detectors technologies. This chapter highlights the need of advanced front-end electronics in the calorimeter and the nature of the environmental constraints. To better understand the functioning of ATLAS calorimeter the second chapter gives a short introduction to the basic principles of calorimetry. Calorimetry is usually very well suited for high energy physics experiments and represents an important part of ATLAS detector. The third chapter gives more details on the implementation of the ATLAS experiment to understand the role of the low voltage power supplies and to set some finer constraints. ATLAS is designed as several layer of subdetectors with different purposes. The low voltage power supplies discussed in this thesis play a fundamental role in the central part of the hadronic calorimeter, powering the front-end electronics and part of the optical reading system. The attention is focused on the tile calorimeter for which details on the architecture, readings and performances are explained, however a short overview is given for all ATLAS subdetectors. The fourth chapter discusses the design and the production of the low voltage power supplies. The supplies consist of different modules (bricks), to provide the different voltages and currents needed inside the detector. The position inside the calorimeter supports implies that the supplies are subjected to the harsh detector environment; furthermore the accessibility is very limited and the bricks are required high reliability. A malfunctioning implies the loss of data from the part of the calorimeter they are powering, causing a non-negligible data loss for the whole detector.

4 The low voltage power supplies installed at the beginning of LHC operation, and, at present, still powering a major part of the calorimeter, have proved to be the weak point of the detector. During 2011 it was observed that the number of malfunctioning in the supplies installed on the detector was increasing with the increase of the colliding beams luminosity. Luminosity is the parameter we use to control the event rate in the LHC and for the research purposes it must be as high as possible; therefore finding a solution to the problem became more and more urgent as the luminosity increased. A new design of the bricks, which addresses this problem as the main point, has been developed to improve the reliability. Prototypes of the new version (v7.5 ) have been installed in the detector and the good results obtained encouraged a big production aimed at replacing all the bricks. During last summer I participated to the checkout procedure testing about 800 bricks out of the 2048 (plus spare) needed. The strict checkout procedure ensures that the brick reaches the design parameter and implements correctly all design functions. Two short term test session verify the main output of the brick and the correct functioning of protection and monitoring circuits. A medium term test session sollecitates the module and points out any early malfunctioning. The first part of chapter four discusses the general design of the brick and the constraints that led to it. Following there is an explanation of the main aspect addressed in the redesign project. Last part of the chapter describes the testing procedure and discusses the results obtained.

5 Contents 1 LHC and ATLAS Introduction LHC Discover possibilities LHC experiments General purpose detectors: ATLAS and CMS Common design Definitions and conventions ATLAS detector design principles Calorimetry Introduction Basic principles of calorimetry Calorimeter performances Electromagnetic Calorimeters Hadronic Calorimeters The ATLAS experiment Introduction ATLAS subsystems Magnet System Inner Detector Calorimeters Muon Spectrometer TileCal Architecture Front-end Electronics Monitoring, testing and calibration Performance Low Voltage Power Supply System Introduction Environmental Constraints Design overview Improvement Motivation for the redesign project Noise Improvements Protection of sensitive part of the circuits Thermal Insulation Performance test in the calorimeter

6 4.5 Production and checkout procedure The test bench and the automated tests Summary of test results Production Status Conclusion 36

7 Chapter 1 LHC and ATLAS 1.1 Introduction The Large Hadron Collider (LHC) [1] at CERN [2] is the bigger and most powerful particle accelerator ever built, it has been realized thanks to the effort of a large international community, the same community that nowadays animates the six big experiments and is eagerly waiting for any discovery of new physics. LHC is colliding proton beams within an entirely new range of energy completely unexplored, way higher than any previous experiment and therefore it allows studying particle interactions in an unexplored range. ATLAS (A Toroidal LHC s ApparatuS) [3], one of the two general purpose detectors installed in LHC, is designed to detect the presence of unknown particles and phenomena generated during the collision. The first results from the study of the collected data have already generated great excitement. In this chapter we will see the main characteristics of LHC and the most awaited discoveries. Then we will look at the common design and problems of the two general purpose detectors installed inside LHC. In the final part we will focus on ATLAS and analyze its characteristics. 1.2 LHC In the mid 1980 s, scientists were thinking of an accelerator more powerful than any technology realized until then, to make further progress in the sudy of matter properties. This idea became reality in 1994 when the CERN s LHC project was approved. The LHC was designed to operate inside the LEP tunnel (a previous electron-positron collider in CERN) and to be the most powerful accelerator ever built, indeed the LHC could boast energy ten times greater than its predecessors and a hundred time greater event rate. Thanks to the effort of people from all over the world, LHC was built and began operation at the end of LHC operates with both protons and lead ions. Protons are preaccelerated in the CERN accelerator system in the order by LINAC, Protosyncrothron (PS) and Super-Protosyncrothron (SPS), and enter the LHC ring with energy about 450 GeV (an overview of CERN accelerator system is given in Figure 1.1). There two beams travel in opposite directions, in two different beam pipes. Pipes are combined in a two-in-one design and their corresponding set of coils are inserted in a unique structure and in a single cryostat [4]. A total of 1232 superconducting NbTb dipole magnets, cooled with superfluid helium at 1.9 K generate a 8.4 T field able to keep the proton beams on the approximately circular orbit. LHC is designed to accelerate the beam up to 7 TeV energy per beam reaching a collision center of mass energy of 14 TeV. Presently the beams are accelerated only up to 4 TeV, the nominal energy of 7 TeV will be reached in 2015 after the upgrade work on the beam line 1

8 Figure 1.1: CERN s accelerator system [4]. will be completed. Protons are grouped in roughly cylindrical bunches few centimeters long and few microns in radius. The time distance between bunches is presently 50 ns and can be shortened up to 25 ns. Each bunch is composed by about protons. The production rate of a particular type of event per second depends on the cross section of the event and on the beam parameters: the number of bunches per beam (n b ), the number of protons (N 1, N 2 ) contained in the colliding bunches, the transversal bunch dimension (A eff ) and the bunch revolution frequence (f). The variable that describes the beam characteristics is termed luminosity and is defined as: 1 N 1 N 2 f L = n b (1.1) 4π A eff The production rate of a process having a cross-section sigma is then expressed in terms of luminosity as: R = σl (1.2) The cross-sections of the processes we are interested in at LHC are very low therefore, in order to have a high enough event rate, the luminosity is required to be very high. This condition however increase the event rate both for interesting events and for non-interesting events (background). The side effect of luminosity increase is therefore the need for a tight trigger selection to record with high efficiency the few interesting events among the countless others. The luminosity has been largely increased from the start of operation thanks to a detailed LHC tuning. At present luminosity is cm 2 s Discover possibilities Little is known for sure about the energy range explored by LHC and this is perhaps the most exciting aspect of the experiments; however lots of hypotheses have been done. Among theoretical prediction the most awaited discovery is surely the Higgs boson.this particle was introduced in the Standard Model to explain how the W and Z gauge bosons could acquire mass while leaving the electro-weak theory gauge invariant. Present theories do not 2

9 predicts any exact mass for the Higgs and do not exclude the existence of more than one Higgs boson. Measurements from previous experiment have set an upper limit to the lighter Higgs of the order of 115 GeV. During the summer (July 4 th ) ATLAS and CMS [5] groups announced the discovery of a new particle with mass 125 GeV, to present, this particle is compatible with the Higgs boson; a complete confirmation of the properties has yet to come. In spite of its good prediction the Standard Model is not expected to be the ultimate physics model, extensions to the model have already been formulated as for example Supersymmetry; this theory predicts new particles with masses below the upper limit of LHC energy, hence interesting results are expected from it too. Leptons and quarks are considered elementary particles in modern physics, but it may not be true; if they are not elementary, LHC is supposed to give signals of their compositeness. There are also open questions waiting for LHC answers, since the LHC reproduces the condition of the early universe, some cosmological question may find answers, like what dark matter is or why do we observe an asimmetry between matter and antimatter LHC experiments LHC hosts at all 6 experiments. The discoveries just depicted will most probably come from the two general purpose detectors ATLAS and CMS (Compact Muon Solenoid). The other experiments have been designed to carry out more specialized studies: ALICE (A Large Ion Collider Experiments) will analyze in detail the products of the Pb-Pb collision, as for example the property of the expected quark-gluon plasma. LHCb instead will be specialized in the physics of B-mesons. TOTEM and LHCf are smaller detectors, connected respectively to CMS and ATLAS, that study the soft collision characteristics. 1.3 General purpose detectors: ATLAS and CMS We repeatedly referred to ATLAS and CMS as general purpose detectors, with this expression we indicate instruments able to detect a wide set of particles in a vast kinematic range, with high enough precision to allow inferring the possible presence of new particles or phenomena. ATLAS and CMS are general purpose particles detectors situated in LHC, they are able to detect electrons, photons, muons, hadrons and jets that are created in the collision or in the decay of particles generated in the collision processes. Covering a large solid angle around the collision point the detectors are also able to infer the presence of neutrinos by missing transverse momentum technique. The fact that both detectors are designed to study the same type of phenomena may appear redundant, on the contrary this characteristic is very useful since it allows immediate cross checks of possible discoveries with independent techniques. The identical purpose and working conditions shape the two detector quite alike, on the contrary the detailed design of each detector and the choice on the magnetic system are so different to give each one a unique personality Common design ATLAS and CMS have a cylindrical shape very long along the beam direction to minimize the angle covered by the beam pipes and the chances of energy escaping that way. Both detectors are composed by several layers of particles detectors, with different roles. The layers are disposed concentrically around the collision point; a set of magnets, very different for the two detectors, completes the design. The inner layer is constituted by a tracker immersed in the magnetic field, which bends 3

10 charged particles providing measurement of charge and momentum. The tracking information allows the reconstruction of collision vertices. The tracking detector is characterized by a fine granularity needed to resolve close by particles in the dense environment. Going outward we find the calorimeters, both detectors have distinct electromagnetic and hadronic sections, with the electromagnetic one closer to the impact point. The calorimeters are designed to provide as hermetic coverage as possible, since any energy leak could be confused with energy carried away by undetected particles such as neutrinos. The required coverage is achieved by instrumenting the region very close to the beam pipes to be able to measure particles up to very low angles and by providing enough material to fully absorb the showers. The latter condition defines the calorimeters dimension computed in radiation and interaction length. Given the materials used in the two detectors the radiation length of the electromagnetic shower goes from 5 mm to 1 cm and for fully absorption are necessary about 25 radiation lengths; for the hadronic shower the interaction lengths is about 15 cm and about 10 interaction length are needed for a full absorption [6] (Radiation and interaction length are typical unit for calorimetry and depend on the calorimeter material, more details on this topic will be given in the next chapter: section ). Muons do not shower in the calorimeter, so they need a dedicated stage; large trackers at the outer radius measure charge and pulse of muons. The magnet system is different for the two detectors: ATLAS choice is one solenoid in the inner part of the detector and three large toroids at the outer radius, CMS on the contrary use only one solenoid at the outer radius of the detector [7]. The particles are identified analyzing their signature in the different parts of the detector, for example electrons are bended by the magnetic field in the inner tracker and then shower in the electromagnetic calorimeter, neutrons do not leave any trace in the tracker and little in the electromagnetic calorimeter but they shower in the hadronic calorimeter. A schematic view of the signature of various particles is shown in Figure 1.2 and summarized in Table 1.1. Figure 1.2: Basic detecting system of ATLAS and CMS detectors system, based on ATLAS image [8]. Neutrinos are detected through missing transverse energy technique. Transveres energy (E T )is a vectorial quantity associated with energy and direction of the impinging particle; in formula, called θ the polar angle of the calorimeter cell hitted by the particle, we have: E T = Esin(Θ) (1.3) 4

11 Inner Electromagnetic Hadronic Muon Tracker Calorimeter Calorimeter Detector e X X γ X p (charged hadrons) X X X n X X ν µ X X X X Table 1.1: Element of detection for the main components of the collision products (summary of information in 1.2). Transverse energy is conserved during the collision. If a neutrino carries away part of the energy, (escaped energy), the vectorial sum of detected transverse energies will not be zero and the neutrino energy will result as missing transverse energy, ET miss. The two detectors are able to analyze products from all the energy spectrum available at LHC, but the high efficiency of the accelerator, while allowing to study a wide spectrum of physics still unexplored, creates at the same time an hard environment for the detectors. Event rate at full luminosity is 10 9 events per second, with bunch intervals of 25 ns, while present data storage ability is 10 2 events per second, luckily not all the events are relevant. ATLAS and CMS experiments are interested only to a small fraction of these events (the ones with creation of new particles) therefore an efficient data selection is necessary while acquiring the data. The system in charge of the selection is called trigger. The trigger system is designed very differently for the two experiments, however the overall conditions which determine trigger selections are the same: pile-up and QCD background. The events we are interested involve directly the proton constituents (quarks and gluons), we call them short distance events, these are about 1/25 of total collision. The remaining is long distance events which involve the proton as a whole. One of the main differences between them is the transverse momentum, p t : short distance events have usually high p t, while long distance events have always little p t. The first requirement to the trigger system is then to cutoff long distance events. The events at high p t are dominated by jet production: after the collision, quarks may hadronize and give birth to collimated fluxes of hadrons called jet; this process can happen through a lot of channels, this is why it dominates the high p t event rate, we refer to these events as multi-jet events. The chances to detect a new particle produced with a low crosssection through a decay channel in two or more jets is very slim and the trigger can be set to give up all these events including the interesting event to privilege clearer channels. On the recorded events another problem arises: the pile-up. During bunch crossing more proton pairs can collide simultaneously due to the high density of particles in the bunches, each collision point is the geometrical vertex of the tracks of the particles produced in the collision. If one of the collision evolves in a relevant event, the detector records all the signatures from the crossing without distinguishing the various collision; the additional events recorded overlap the signature from the interesting event and are called pile-up events. Pileup events, can be distinguished from the relevant event only after vertex reconstruction Definitions and conventions Before continuing with ATLAS analysis it is good to familiarize with few quantities that will be used: whenever speaking about coordinates, we consider the origin of the axis in the collision point. The z axis lies on the tangent to the LHC s ring and its positive orientation is the counterclockwise, the x axis points toward the center of the ring and the y positive direction is upward (the three axis created a right-hand triplet). Whenever we will refer to 5

12 p T, in the following, we consider it defined on the x, y plane. Very often we will use also radial coordinates that are obtain as: R = x 2 + y 2 φ = arctan y x θ = arccos z R 2 + z 2 (1.4) In substitution of the angle θ will be used pseudorapidity (η), this quantity represents θ, but distribution of η are Lorentz invariant along z for ultrarelativistic particles, the connection between the angle and pseudorapidity is: ( θ η = ln tan 2) (1.5) 1.4 ATLAS detector design principles Figure 1.3: Schematics of ATLAS detector [8]. There were two main aspects leading the design of ATLAS: the characteristic of the events to be measured and the different techniques available to implement each subdetector. The main criteria that have driven the ATLAS detector are discussed in detail in the ATLAS Technical Design Report and are here summarized [9]: very good electromagnetic calorimetry for electron and photon identification and measurement, complemented by full-coverage hadronic calorimetry for accurate jet and missing transverse energy (ET miss ) measurement; high-precision muon momentum measurement, with the capability to guarantee accurate measurement at the highest luminosity using the external muon spectrometer alone; 6

13 large acceptance in pseudorapidity (η) with almost full azimuthal angle (φ) coverage everywhere; efficient tracking also to be used for efficient track reconstruction and heavy-flavour and τ identification; triggering and measurement of particles at low p T thresholds, providing high efficiencies for most physics processes of interest at LHC. The detector is to give redundant information, to allow dealing with the harsh environment of LHC. As seen in the previous section the number of particles impinging on the detector varies largely as a function of pseudorapdity, this characteristic influences the whole design of the detector. Each subsystem is divided in few geometrical areas where different detection techniques are used in order to obtain the best performance while mantaining a sufficient radiation resistance. Roughly the detector is assembled in three parts barrel, end-cap and forward, covering small, intermediate and large η regions (Table 1.2). Detector Section Pseudorapidity region Barrel η < 1 End-caps 1 < η < 2 Forward 2 < η < 4.5 Table 1.2: Rough η regions division for ATLAS detector [7]. 7

14 Chapter 2 Calorimetry 2.1 Introduction A fundamental technique for the detection of a wide range of particles both charged and neutral is calorimetry. Calorimetry is a very adaptive detection principle used in particle physics. With this method the particle is fully absorbed and an active material generates a signal proportional to the particle energy. Calorimetry made its first appearance in cosmic rays study [10], at first it was just an empirical technique which provided good results, with time we started to understand the physics beyond the phenomena observed and we became able to control and improve the technique; meanwhile the development of accelerators requested calorimeter with higher performance. Nowadays, calorimeters are an essential component in all high energy physics detectors. In this chapter we will discuss the main principles of calorimetry and the key variables to assess the calorimeter performance. 2.2 Basic principles of calorimetry Modern calorimeters are very adaptive and they can be specialized: we usually distinguish electromagnetic calorimeters, used to detect mainly electrons and photons, and hadronic calorimeters, used to detect hadrons and jets. Another characterization of calorimeters can be done upon their architecture: homogeneous and sampling calorimeters. The first ones consist completely of active material while the second ones are composed of layers of active material alternated with absorber. The incident particle, interacting with the calorimeter material in a manner peculiar to the particular particle type, generates few secondary particles with energy lower than that of the primary particle. These products in turn produce other particles by the same mechanism, thus giving rise to a cascade (shower) of particles with progressively degraded energies. The number of particles in the shower increases until the energy of the particles falls below a critical energy, thereafter energy is dissipated mainly by ionization and excitation rather than by the generation of other particles. As final result we obtain a shower of particles whose length depends on the incident particle energy and on the calorimeter material. One of calorimetry s main features is the ability to detect and distinguish a wide range of particles, including neutral particles such as neutron, through strong interaction, or π 0, through its decays. The energy deposited in the active material is converted into light or charge, producing a signal which is proportional to the primary interacting particle energy. In order to resolve the energy deposited by different primary particles the calorimeters are segmented in cells along the transversal direction: each cell provides the signal integrated on a small geometrical region. The smaller are the cells the smaller is the possibility that the signal of a single cell 8

15 is due to two particles superposed. The dimension of the cells depends on the calorimeter requirement, we refer to this characteristic as granularity of the calorimeter (with reference to photography). Some calorimeters have also a longitudinal segmentation, this allows to take more sampling at various shower depths, thus to follow the shower development. The additional information allows us to infer the direction of the incident particle. A great advantage of calorimeters is that the fractional resolution improves with energy, in particular, since the number of particles in the shower is roughly proportional to the energy of the incident particle, N E, we have that: E E N N N N 1 N 1 E (2.1) At the beginning of the shower each particle interacts with the material generating two or more particles, with a rough estimation we can consider the growth of the shower exponentially and we can suppose that each particle receives approximately the same amount of energy. With this approximation the shower length increases only logarithmically with the increase of energy and obviously the calorimeter s dimension increases on the same scale, for this reason calorimeters are space and cost effective. As a last consideration about calorimeters advantages we must say that some calorimeters can produce very fast signals and therefore they can be used to generate a fast signal for low level triggering stages Calorimeter performances The performance of a calorimeter is normally estimated using two variables: the energy response and the energy resolution. The response is the fraction of energy detected over total energy of the particle: response = energy detected energy of the incident particle (2.2) If we consider the response as a function of the energy of the incident particle, we call it calorimeter linearity: we say that a calorimeter is linear if the response is independent of the energy of the incident particle. As said before if we consider only the statistical fluctuations of the number of particles composing the shower, the fractional resolution improves with the square root of the incident particle energy. A more precise formula for the resolution can be written as sum in quadrature of three terms: (Cfluct ) E 2 E = + E ( Cinstr E ) 2 + C 2 const (2.3) where C fluct / E is the statistical term, C instr /E is the noise term and C const takes into account non-uniformities of the calorimeter. The statistical term is normally the dominant term in sampling calorimeter with C fluct having values of the order of 10% in electromagnetic calorimeter and up to 100% in hadronic calorimeters. The noise term is highly suppressed by the 1/E term and therefore is usually negligible at high energy. 9

16 2.3 Electromagnetic Calorimeters Electromagnetic calorimeters are designed to detect electrons and photons. The interaction of these particles are well described in quantum electrodynamics (QED) [11]. The interactions with the material for electrons and photons result in a quite different behavior of these two particles, while the energy of electrons is degraded without destroying the electrons, photons are typically transformed in other particles. For this reason while the interaction of electrons is described using the energy loss per unit length, the interaction of photons is described using the cross-sections. The different processes that contribute to the energy degradation of electrons interacting with the material are shown in Figure 2.1 (left) as a function of the electron energy. The cross sections of the interaction of photons are shown in Figure 2.1 (right) as a function of the photon energy. The graphics shown in Figure 2.1 are obtained for electrons and photons interacting with lead. Figure 2.1: LEFT: contribution of different processes to the energy loss per unit length of electron interacting in lead as a function of the electron energy [10]. RIGHT: cross sections of different processes for photon interacting in lead as a function of the photon energy [10]. Electrons of energies larger than about 10 MeV loose energy predominantly by bremsstrahlung [10] due to the crossing of the nuclear field while photon interactions in this energy range, produce mainly electron-positron pairs. Secondary photons, electrons and positrons, if they have high enough energy, can interact through the same processes generating the shower, until they reach the critical energy. Below the critical energy the main interaction processes are ionization and thermal excitation for electrons and Compton scattering and photoelectric effect for photons. When considering these phenomena and more in general calorimeter dimension a convenient scale is provided by the interaction length X 0. For electrons one radiation length describes the rate at which electrons lose energy by bremsstrahlung, in fact one radiation length is the average distance that an electron has to cross in a material to reduce its energy to 1/e of the initial energy: E(x) = E 0 e x X 0 (2.4) Similarly, for a photon beam the radiation length describes the rate at which the beam intensity is reduced by pair production. In this case one radiation length is 9/7 of the average amount of matter after which the photon beam intensity is reduced to 1/e of the initial intensity. x 9 I(x) = I 0 e 7 X 0 (2.5) 10

17 The radiation length depends on the characteristic of the material and can be estimated as [10] X ( Z(Z + 1) ln 287/ )g cm 2 A (2.6) Z where A and Z are the mass and the atomic numbers of the medium respectively. There are two different definitions for critical energy for electrons: it can be considered as the energy at which the energy losses from bremsstrahlung are the same as energy losses from ionization, or it can be the energy for which ionization losses per radiation length are equal to the electron energy; these definitions are equivalent in the approximation: ( ) de E (2.7) dx bremsstrahlung X 0 where X 0 is the radiation length. The critical energy (ɛ) depends on the characterisitics of the material and can be approximated by [10]: ɛ MeV (solids) ɛ MeV (gases) (2.8) Z Z Figure 2.1 shows that ɛ 7 MeV in lead. Equations (2.4) and (2.5) show that the physical scale over which a shower develops is similar for incident electrons and photons, and that it is independent of the material type if expressed in terms of X 0. Therefore electromagnetic showers can be described in a universal way by using simple functions of the radiation length. At LHC the typical longitudinal size of electromagnetic calorimeters is about 25 X 0, for this dimension the average energy that leacks beyond the material is smaller than 1%. 2.4 Hadronic Calorimeters By analogy of electromagnetic showers, the energy degradation of hadrons proceeds through an increasing number of (mostly) strong interactions with the calorimeter material. However the detailed description of the development of an hadron shower is much more complex than that of an electromagnetic shower. A simplified description of the hadronic interactions can be given dividing the processes in two classes of effects. In the first class we have processes where the incident hadrons, with energy at the Gev scale, produce secondary hadrons with a mean energy of the same order and a mean free path ( interaction length ) of λ = 35 A 1/3 g cm 2 [10]. In the second class we have collisions with the medium nuclei that produce different processes as spallation, nucleon evaporation etc., leaving behind particles with energy in the MeV scale and consuming a relevant part of the initial energy. To better understand the particles generation process we can look at Figure 2.2 where the average energy spectra of the major types of particles produced by 100 GeV protons interacting with lead are shown. The number of particles produced is proportional to the energy of the incident particle, the trend is quite linear, Figure 2.3 shows this trend for the neutrons component of the shower for different energies of the incident hadrons. The deviation is due to the presence of neutral pions in the shower; this particular kind of particle is produced by strong processes, but decays 11

18 Figure 2.2: Spectra of shower components from 100 GeV protons in lead [10]. Figure 2.3: Neutrons produced by protons on lead. With increasing enegy of the incident particle, the hadronic component is reduced relatively to the electromagnetic component [10]. electromagnetically (before having a chance to interact hadronically); therefore it transfers the energy from a further hadrons production to an electromagnetic shower. As the energy of the primary hadron increases so do the number of hadronic interactions and the number of π 0 produced, leading to a larger fraction of the electromagnetic component in the cascade. Assuming a cutoff value for further hadronic production E GeV, the fraction of hadronic fraction of a shower induced by a hadron of energy E can be described 12

19 by: ( ) E k F h = where k = ln α E0 ln m (2.9) where m is the average multiplicity of the fast hadrons produced in a hadronic collision and α the fraction of hadrons not decaying electromagnetically; k is of order k 0.2 therefore F h 0.5(0.3) for primary hadrons of E=100(1000) GeV [10]. The hadronic fraction decrease with increasing energy. The fraction F h also depends on the type of the incident hadrons, usually a baryon induced shower will contain little less π 0 than a meson induced shower. Nuclear reactions can produce neutron and photons as well. In most cases they are slow neutrons and delayed photons; the energy they carry, as well as the binding energy from isotopes formation, is poorly detected by the calorimeter, if detected at all; usually we refer to this as invisible energy. The invisible energy has event by event fluctuation and influences the response of the calorimeter; there are few easy steps which will help us understand this influence. Let η e be the detection efficiency of a signal and η h the efficiency for an hypothetical shower composed only of the hadronic component. Then the visible energy Evis π for a shower induced from a pion of energy E can be written as: E π vis = η e F π 0E + η h F h E (2.10) with the constraint F h + F π 0 = 1 ( E vis = η e F π 0 + η ) h F h E (2.11) η e The observable portion, i.e., the visible signals induced by electromagnetic (E e vis ) and hadronic showers, usually denoted e/π, is therefore: E π vis E e vis ( ) e 1 ( = 1 π 1 + η h η e ) F h (2.12) In general η e η h and the event to event fluctuation, in the fractions F e and F π 0, dominates over the detector fluctuation and greatly influences the calorimeter response making the calorimeter non linear and degrading the energy resolution. Calorimeters that have η e η h are said to be non compensated. Calorimeter compensation can be achieved with special techniques like using high Z absorber or using active material rich in hydrogen [12]. In sampling calorimeters, calorimeter compensation may also be obtained via software exploiting the information of the shower shape. 13

20 Chapter 3 The ATLAS experiment 3.1 Introduction We saw that ATLAS is composed by several subsystem disposed concentrically around the interaction point. Each subsytem is designed to obtain the best performance in the high pileup and high radiation environment of LHC. The easier way to describe the complex ATLAS architecture is to proceed outward starting from the interaction point. Before describing each subsystem we describe the magnetic system that plays a fundamental role for the measurement of the transverse momenta of charged particles. The TileCal system is described more in detail since it is the subject of this thesis. An overall view of the ATLAS experiment is shown in Figure ATLAS subsystems Figure 3.1: Schematics of ATLAS detector with indication of main subdetector division [8] Magnet System The magnet system is composed of a central solenoid (CS) situated in front of the electromagnetic calorimeter and of three air-core toroids situated on the external part of ATLAS detector, one in the barrel section (BT) and two at the end-caps (ECT). In the transition region, the magnetic field is generated by the superposition of the BT and ECT fields. The central solenoid provides a nominal 2 T magnetic field to the inner detector region, with 14

21 a peak of 2.6 T in proximity to the coil. It is designed to minimize the material and it shares a vacuum vessel with the electromagnetic calorimeter thereby eliminating two walls. The three toroids consist of eight coils assembled radially and symmetrically around the beam axis. In the BT each coil is housed in a different cryostat (that take up the forces between coils), and the toroidal structure is completed by the linking elements, which provide mechanical stability. A cryogenic ring, which connects the eight cryostats to a service one, provides connections to the power supply, to the helium refrigerator, to the vacuum system and to the control system. In the ECT instead the eight coils are assembled in one single cryostat and the internal forces are taken by a cold supporting structure between the elements. The coils are positioned on a rail to facilitate the access to the inner part of the detector. Helium flows at 4.5 K and it is forced through tubes welded on the casing of the windings and indirectly cools the magnets. In addition to this, the CS is cooled via a dewar coupled to the refrigerator, while the BT and the ECT have a cold helium pumps. The cooling power is common for the four magnets Inner Detector The inner detector or inner tracker is the closer detector to the interaction point, it must have high granularity to cope with the high density of tracks and, at the same time, it must consist of little material to minimize the number of interaction length placed in front of the calorimeter. This is realized using two technologies, closer to the interaction point, the highest granularity is achieved with 8 silicon microstrip (SCT) layers and 3 silicon pixel layers, while at larger radius a detector based on straw tube trackers provides a large number of tracking points with little material. The combination of these two techniques gives very efficient pattern recognition and high precision. The pixel detector consists of three barrel placed at 4, 10 and 13 cm from the interaction point and five disks on each end cap at distances between 11 and 20 cm; the system is designed to be modular and uses identical modules on the barrel and in the end-caps. In the barrel a small angle stereo between microstrip layers provides the z measurement. Endcap SCT are mounted up in three rings onto nine wheels, which are interconnected by a space frame. The inner detector is responsible for tracking and vertex reconstruction. The very high spatial resolution needed for these measurement requires not only a challenging granularity, but also a very good stability of the system to any distortion of movements. In order to obtain the needed stability the detector is built with material with little thermal expansion coefficient moreover cold operation and removal of the heat are used as well to comply with the requirement. Transition Radiation Tracker (TRT), in the transition regions, uses straw detectors which can operate at high rate thanks to the small diameters and the isolation of the sense wires within individual gas volumes. Using a radiator between straws it is possible to identify electrons through the emission of transition-radiation photons. This technique is radiation hard and allows a large number of measurements on every track Calorimeters ATLAS calorimeter is divided in two sections one electromagnetic closer to the impact point and one hadronic further from the impact point. While the electromagnetic architecture is quite similar in the barrel and end caps regions, the hadronic calorimeter segments have very different architectures. In the hadronic section we usually consider also the forward calorimeter which actually is a generic calorimeter. 15

22 Calorimeter Region Electromagnetic Barrel Calorimeter η < 1.5 Electromagnetic End-cap Calorimeter 1.5 < η < 3.2 Hadronic Barrel Calorimeter η < 1.7 Hadronic End-cap Calorimeter 1.5 < η < 3.2 Forward Calorimeter 3.1 < η < 4.9 Table 3.1: Resume of ATLAS calorimeter segmentation [9]. Electromagnetic Calorimeter The electromagnetic calorimeter is a liquid argon sampling calorimeter. It uses lead plates as absorber and Kapton electrodes, which are disposed in an accordion geometry with φ symmetry; this disposition is very characteristic of ATLAS. The material in front of the calorimeter is about 2.3 X 0 at η = 0 and increases with the pseudorapidity due to the particle incidence angle. In the region η < 1, 8 where most precise measurement are required, a presampling calorimeter, located in front of the electromagnetic calorimeter, is used to correct the energy loss in the upstream material; in the transition region where the thickness of the material in front of the calorimeter reaches 7 X 0 the presampling calorimeter is complemented by a scintillator slab inserted in the crack between the barrel and end cap calorimeter sections. The barrel section is composed by two identical half barrel separated by a 6 mm gap at z = 0, while the end-caps calorimeters consist of two coaxial wheels (1.375 < η < 2.5, 2, 5 < η < 3.2). The total thickness of the EM calorimeter is > 24 X 0 in the barrel and > 26 X 0 in the end-caps. Hadronic Calorimeters ATLAS hadronic calorimeter covers the region η < 4, 9 enough to provide excellent missing transverse energy resolution. The large angle is possible thanks to a triple differentiation: forward calorimeter, end cap calorimeter and tile calorimeter. The hadronic calorimeter is the external layer of ATLAS calorimeter, thus it must be as absorbing as possible and reduce the hits on the muon spectrometer, the radiation length required for this purpose is about 10 radiation length. Total dimension of hadronic calorimeter corresponds to about 9 λ of which 7.4 λ are instrumented and about 1.8 λ instrumented is provided in the electromagnetic calorimeter. However the total material transversed corresponds to 11 λ. The LAr forward calorimeter (FCAL) covers the region closer to the beam pipe, 3.2 < η < 4.9, the calorimeter is supposed to provide at least 9λ of material in as little space as possible; the fact that it is moved back from the electromagnetic end-cap to minimize the albedo, further reduced this space; the solution is high density material. The forward calorimeter has been realized as three metal matrices on each side of the detector, one is filled with copper rods and tubes while in the other two the rods and tubes are from tungsten, the active material (liquid Argon) fills the gaps. The use of tungsten was rather new at the time of FCAL designing. The LAr Hadronic End-Cap Calorimeter (HEC) covers the region 1.5 < η < 3.2. It is composed by two copper wheels with outer radius 2.03 m, the outer one with a thickness double of the inner and each wheel is divided in 32 modules. Parallel to the plate in the gap there are three electrodes, one for the read-out and two for high voltage carriers, the gap is then filled with liquid argon as active material. The Tile Calorimeter(TileCAL) will be discussed in the next section. 16

23 3.2.4 Muon Spectrometer High precision tracking chambers detect muons when they crossed the large air-core toroidal magnets at the outer radius of ATLAS detector. In this region, the field, produced by the barrel toroid ( η < 1), the end-cap toroids (1.4 < η < 2.5) or a superposition of the two ( 1 < η < 1.4), is perpendicular to the particles direction almost everywhere and bends the particles allowing momentum and energy measurements. Monitored Drift Tubes (MDTs) effectuate the measurement for a long range of η, but at large pseudorapidities and close to the interaction point Cathode Strips Chambers(CSCs) with higher granularity are preferred due to the bigger amount of tracks in that regions. The geometric disposition is as for the other detectors in concentric cylinders in the barrel and parallel plates in the end-caps; precision measurement are effectuated along the z direction in the barrel, and R direction in the transition and end-cap region, however measuring only one direction is reductive for ATLAS purposes; the trigger system complements this measurement with a second direction orthogonal to that of the high precision chambers. In fact the trigger system is composed by Resistive Plate Chambers (RPCs) in the barrel and Thin Gap Chambers (TGCs) in the end-caps, which can measure the energy in one direction as well as identify bunch crossing, and provide a trigger with well defined p T cut-off in moderate fields. The background that affect the muon spectrometer instrumentation is mainly composed by n and γ originated in secondary interactions in the various elements. 3.3 TileCal The central part of ATLAS hadronic calorimeter( η < 1.7) has be implemented as a sampling non compensating calorimeter with the use of scintillating tiles as active material, from which the name tile calorimeter or shortly TileCal. This choice while achieving the physics goals, makes the calorimeter cost effective and facilitates the modular architecture, which allowed the calorimeter construction to be distributed between several institution. The tile calorimeter uses steel as absorber, the modules are mounted on external mechanical support (girders). The overall structure is self supporting and strong enough to provide direct or indirect mechanical support to all others subdetectors of ATLAS, with the exception of the muon detector and toroidal magnets Architecture TileCal is divided in two sections: the barrel, a cylinder which covers the region η < 1.0 and the extended barrel, two cylinders at 0.8 < η < 1.7. Between the two regions a gap allows the passage of services for the electromagnetic calorimeter and the inner detector. All the sections share the same internal structure, included the orientation of tiles. Inside Tilecal, unlike what happens in the inner detector and electromagnetic calorimeter, the tile orientation does not depend on the initial particle direction; in fact most of the particles have already started showering in the inner material of the detector and the shower particle directions are spread over a very large angle. The choice for orientation was then made to make the calorimeter as hermetic as possible, beam tests proved that the best disposition for this purpose is a radial disposition with a stagger in depth. The thickness of a single scintillating tile in the z direction is 3 mm, while the lateral dimension varies with the radius. The tiles were produced with an innovative method appositely designed. The scintillating tiles are immersed in the steel which works as absorber. The steel has a laminar structure oriented in the radial direction, full plates of steel in the shape of wedges, masters, are separated from each other by small trapezoidal steel plates, spacers, large almost 17

24 as the masters and spread along the bigger plates. A gap delimited by two masters and two spacers creates a sort of pocket in the structure and host a scintillating tile. In the Figure 3.2 is shown the shape of the final wedge, with the girders on top. In the enlarged detail on the right the pocket for the insertion of the tiles are represented as black marked lines. Figure 3.2: Schematics of one of TileCal modules. The module has a wedge shape and it is attached at the girders (on the top of the image) which supports the calorimeter and hosts the calorimeter electronics. Coupling of the wave length fiber to the tiles and the photo multipliers is shown. In the enlarged detail is represented the steel structure, the black marked segments represent the pocket for the insertion of the scintillating tiles. The little circles are the hole for the passage of the Cs source [10]. The spacers start at 1.5 mm from the radial edge of the master plates to allow the passage of optical fibers for the reading [13]. The scintillating tiles before insertion in the absorber structure are enclosed in appositely designed Tyvek R sleeves which protect them from damaging. Tyvek is a very resistant material with consistency similar to paper and little costs, thus it is well suited for the purpose. When the Tivek was applied to the tiles it was observed that the light at the edge was brighter than in the center of the tile due to high reflectivity of the material. The non uniformity of the light inside the tiles caused errors in the reading, thus it was decided to apply a mask to the Tyveck to reduce the material reflectivity. Wave length shifting fibers (WLS) in polystyrene are coupled with each of the radial edge of the scintillating tile and routed in small slots in the absorber material in the radial direction; the fibers are coupled directly with the tiles and double cladded for their entire length to exclude other signal, the cladding also increase the light yield and the durability of the fiber. The blue-green light produced by the tiles is shifted to wavelength of 490nm, with the use of a fast dye mixed directly in the fiber material. Since the dye is distributed uniformly in the fibers, the shifting on the emitted light depends on the length of the fiber. Fibers from the same cell are combined into a bundle and routed to the same photo-multiplier tube (PMT), the photomultiplier is located as close as possible to the geometric center of the cell to obtain an homogeneous response from signals at the same depth. A schematic overview of the fiber coupling to the cells and PMT is given in Figure 3.2. The cell transverse dimension corresponds to the lateral spread of the hadronic shower (λ 23 cm). The longitudinal dimension varies with the radius: close to the internal and external radius of TileCal are thin layers while in the middle thick ones. In this disposition the first thin layer allows to correct for energy losses in the cryostat in front of the calorimeter by correlating this response to 18

25 that in the last layer of the electromagnetic calorimeter. The central layer absorbs the bulk of the shower. The last thin layer capture the residual of the shower and minimize losses. Each layer is large enough to provide an integrated muon signal which can be used for trigger and records the passage of muons non-reaching the muon spectrometer. The PMTs transmit the fiber signal to the front-end electronics for data handling and storage of the event Front-end Electronics TileCal front-end electronics is located at the outer radius of the calorimeter in an extension of the girders called finger, more precisely in a removable part, the drawer, which slids in the finger during operation. In the drawer are also located low voltage power supplies, cooling and readout cables. This position has the advantages of the finger partially shielding these components from the return magnetic field of the central solenoid and from the residual products of the collision and provides as well physically protection of the cables. Nevertheless the components in the drawer are under the influx of the fields from the toroidal magnets and are transversed by a moderate amount of radiation, in addition the drawer has two main issues: little space and limited accessibility. The front-end electronics processes the signals generated by the photomultiplier, provides the circuits that handle the calibration system with a radioactive source and generates the trigger signals. The front end electronics is composed by several subsystems: the 3-in-1 card; the motherboard system; the digitizer boards; the optical interface board. The 3-in-1 card removes shape fluctuations from the PMT signals and produces a standard signal shape for all TileCal channels. In order to cover the full dynamic range (30 MeV - 2 TeV) with the required resolution the signal is amplified by two parallel electronic circuits (low and high gain) having amplifications with a ratio of 1:64. The amplified signals are then routed to the differential drivers which send them to the digitizer. At the same time, the low gain signal provides a differential fast trigger signal sent to the trigger sum boards mounted on the motherboards. The cable length regulates the fine timing for the analog sums. The 3-in-1 system also provides the circuitry for a charge injection system which allows to send pulses of pre-defines amplitude and timing, directly to the shaping network, allowing to monitor the linearity of the amplification system. The motherboard system controls up to 48 3-in-1 cards in the drawer, its primary function is decoding the commands from the trigger, timing and control (TTC) and sending appropriate instructions to the 3-in-1 boards. The TTC is extensively used in control of LHC machine. Commands are received via optical cable and decoded on the first motherboard. They are then sent on a serial differential bus to 3-in-1 cards. They can be sent to individual 3-in-1 cards or in any subset of cards. The digitizer boards process the signal from the high gain and low gain channels of the 3-in-1 cards and every 25 ns digitize it by 10-bit ADC. Normally each event corresponds to 7 digitized samples. After digitization the high gain signal is transmitted to the next stage if none of the digitized samples are saturated otherwise the low gain signal is transmitted. The optical interface card receives data from the digitizers, realigns them to a common clock and packs them into 32 bits words for transmission to the readout driver crate in the counting room. The interface card memory holds 16 events at 7 samples. 19

26 3.3.3 Monitoring, testing and calibration TileCal readout can be calibrated, tested and monitored, at various stages, several components and techniques are used to develop this functions obtaining a redundancy in the procedures which grants reliability and and the possibility of crosschecks. The main systems used for the purpose are: charge injection system; radioactive cesium system; laser system. The charge injection system is situated in the 3-in-1 board and allows to monitor the linearity of the front-end electronics and of the gain circuits for the PMT signal. With this method we can precisely calibrate the readout of each ADC channel from ADC counts to picocoulombs (injected charge). The radioactive cesium system is composed by three 137 Cs sources which are moved, with an hydraulic system, in tubes which run along the calorimeter; the sources excite individually each of the tiles with a known signal allowing to test the tile response, the tile WLS coupling and the integrity of the fibers. The response from the various tiles can be equalized by regulating the PMT voltage. The laser system transports fast laser pulses to the PMT through clear optical fiber. This mechanism simulated the activation of one of the tile. In this way it is possible to check the response of PMT and the readout chain processing time Performance Figure 3.3: Energy linearity (up) and resolution (down) at the electromagnetic scale, for pions impinging on the calroimeter at η = 0, 35. Both plots show experimental data (empty circles) and Monte Carlo simulation (filled circles) [14]. To evaluate the TileCal performance, as explained in chapter 2, we can look at the energy measurements of jets in terms of linearity and resolution. TileCal is required to achieve linearity within a few percent, in the energy range between few hundreds MeV (corresponding to the minimum energy deposited in the calorimeter by ionization processes) up to TeV. Linearity is important in the jet detection and influences strongly the sensitivity to new physics. The resolution is demanded to have a statistical fluctuation term of the order of 50%/ E but the other terms should be reduced to about 3%. This level of resolution has been set to detect the decay W jj with a precision of 10%. The performance on the jet energy measurement just exhibited are obtained taking into account both the electromagnetic and hadronic calorimeter. TileCal overall performance have been assessed with the use of pion 20

27 beams with energy between 20 and 180 GeV. In Figure 3.3 the linearity and resolution results from the tests have been compared with Monte Carlo simulation. The plot taken at the electromagnetic scale shows clearly the non compensating behavior. 21

28 Chapter 4 Low Voltage Power Supply System 4.1 Introduction We have already seen why TileCal needs integrated electronics. It is also important to note that a malfunctioning in the electronic components of a module, would undoubtedly involve a severe data loss from the module, that is a considerable data loss for the whole detector. With this premise it becomes interesting to see how the electronics is powered and how power supplies can be built to work reliably in the detector environment. There are two power stages, the first one is located in an underground hall reserved for AT- LAS services and electronics (USA15 or Underground Service ATLAS cavern), here the high power supplies (HPS1) convert 230 V AC in 200 V DC, from there power reaches the ATLAS cavern and it is distributed first to the distribution boxes (one per 4 drawers), then to the 256 low voltage power supplies (LVPS boxes) [15]. Each LVPS box serves one drawer and consists of 8 modules, called bricks, controlled by an embedded local mother board (ELMB) which also serves as an interface with the exterior world for control and monitoring signals. The second conversion stage is implemented in the bricks where 200 V DC is converted in the low voltages DC levels needed by the front end electronics, digitizers and motherboard, and by the HV distribution system which powers the photomultipliers. A scheme arrengement of the power distribution system is shown in Figure 4.1) Considerable constraints to the LVPS boxes come from the detector environment, as repeated several times in previous chapters. Working inside TileCal means to cope with high radiation levels, to be immersed in the Barrel Toroid s magnetic field and to have very little space available. Despite these constraints LVPS are designed to use only commercial off-the-shelf components. During 2011 it was showed that spontaneous failures in the bricks increased with the increasing of luminosity. This phenomena compromises the data taking from the module for few minutes, resulting in a severe data loss for the calorimeter and hence the ATLAS detector. With the progressive increase of luminosity obtained at LHC, there was the urge to find a solution; a redesign project already in place at that time, took this as the main point. The prototypes (v7.3) of the bricks showed a huge improvement; at present the final version of this project (v7.5) is in the final production phase and will be installed in the detector during Last summer I was in Argonne National Laboratory and I cooperated to the brick checkout procedures testing about 800 bricks. In this chapter we will discuss the general design of the LVPS bricks, starting with the environmental constraints of the detector and how they influenced the design. Then we will describe the improvement of the new version of the bricks. The central part of this chapter will be about the production phases and especially on checkout procedure done in Argonne National Laboratory (ANL). In the final part a short overview of the project status will be 22

29 Figure 4.1: Power distribution system for ATLAS [16]. presented. 4.2 Environmental Constraints LVPS boxes are located in the outer part of the drawer and have a particular metallic shielding, to reduce the amount of radiation received and to minimize the influence of the magnetic field, however the shielding does not provide total protection. The residual magnetic field is 0.02 T, enough to have an impact on components such as the power transformers and ferrite chokes. The LVPS boxes are also subjected to the radiation from the calorimeter, in order to better define the radiation constraint for the new design the computation of the amount of radiation received per 10 years LHC runs at full luminosity has been reconsidered. The radiation is computed in terms of total ionizing dose (TID), non ionizing energy loss (NIEL) and single event effect (SEE) and the results are shown in Table 4.1. The three last columns show the radiation dose assumed in 2005 for the present LVPS design (2005 requirements), the survival limit dose and the safety factor assumed for the new design Survival Safety requirements limit factor TID (Grays) NIEL n/cm 2 SEE p/cm 2 Table 4.1: LVPS estimated radiation doses [17]. 23

30 To cope with these radiation levels over long periods, radiation-resistant components were chosen; large scale tests were effectuated to identify the better ones [15]. The metallic shielding while protecting the bricks, encloses them making it hard to dissipate the heat, in addition the small dimensions of the bricks imply that the heat generators are very dense; in such conditions a good cooling system is required. A water cooled plate runs in the middle of the LVPS box, coupling with the plate is provided by ceramic posts strategically placed across the bottom side of each brick. From these conditions it was chosen to use cheaper switching power supplies instead of linear supplies, because the latter would not work properly in the magnetic field and would have bigger dimension that does not fit in the tight space of the drawer. In addition switching supplies are more efficient and, as a consequence, have a smaller heat dissipation. The LVPS s maintenance is difficult due to the strict space constraints and little accessibility to the detector, thus reliability and stability of operations are a severe requirements. To adjust to the output voltage changes, due to natural deterioration, an external trimming for the output voltage was devised by the use of a reference voltage. 4.3 Design overview Switching power supplies are objects able to produce a DC signals with reduced voltages with respect to the input signal while maintaining the power dissipation very low. At the heart of the converter is the high frequency inverter section, where the supply input is chopped at very high frequencies (in the order of khz) to produce the drive pulse. This pulse is then filtered and smoothed to produce the desired DC output. The output characteristic are defined by the frequency and duty cycle of the drive pulse. In the TileCal each brick receives 200 V low current and converts them to the low voltages medium currents needed in the drawer, the requests for the different types of bricks are shown in table 4.2. Brick type Output Voltage Output Current +3DIG V 4.74 A +5DIG +5.3 V 5.6 A -5MB -5.3 V 5.6 A +5MB +5.4 V 11.1 A +15MB V 0.45 A -15MB V 0.3 A +5HV +5.0 V 0.18 A +15HV +14.5V 0.29 A Table 4.2: Voltage and current requests for each brick [17]. All bricks share a common design and different design requirements are satisfied by changing 49 components (personality components) over about 250 components. The basic topology of the brick is a transformer-coupled buck converter. A block diagram of the circuit is shown in Figure 4.2. The heart of the design is the LT1681 controller chip [18]. It is a dual transistor synchronous forward controller able to produce the drive pulse at the frequency of 300 KHz with an output duty factor that can vary from a few percent up to a maximum of 45%. The pulse width is controlled by a feedback circuit based on the values of the output voltage of the brick. This input permits us to ensure continuous-mode operation at the nominal voltages and currents. The signal from the LT1681 is sent to the FET Drivers, IR2110 [19]. These are transistor drivers that have sufficient current and voltage drive to drive the high-side 24

31 Figure 4.2: Block Diagram of the LVPS brick [17]. and low-side power Metal Oxide Semiconductor Field Effect Transistors (MOSFFETs). Both the high-side and low-side transistors turn on and conduct for the duration that the output clock is in the high state, and both are in the off state when the clock is low. When the MOSFETs conduct, current flows through the primary windings of the transformer, which transfers energy to the secondary windings. The transformer has different voltage ratios for 5 V and 15 V bricks. A buck converter [20] is implemented on the secondary side of the transformer and converts the signal to a constant voltage for the output. The output side also contains an additional LC stage for noise filtering. Voltage feedback, for controlling the output voltage, is provided by the Avago opto-isolators HCPL-7800 [21]. This component processes the electrical signal converting it into digital signal, at the input stage, and transmitting it through pulsed light signal to the output where it is converted back to an electrical analog signal. The signal is transmitted mantaining the input and output electrically insulated from each other. The design also incorporates two shunt resistors for measuring the output current, the voltage for which is also fed back using an opto-isolator. The value of the output voltage is controlled by a reference voltage that comes from the central controller in the LVPS box, the ELMB. The brick has three types of protection circuits built-in as part of the design, over voltage protection (OVP) and over current protection (OCP) are on the primary side and are configured specifically for each particular type of brick. The third one is over temperature protection, which monitors the temperatures of the low-side transistor on the primary side. This circuit is integrated in the LT1681 design. When one protection circuit is triggered an off signal is sent to the LT1681 which stops the brick immediately. Several monitor circuits are implemented on the brick, these circuits send the information to the ELMB that in case of need can command a shutdown. In particular the ELMB monitors two temperatures on the primary side of the brick both close to the input of over temperature protection of the LT1681 and the output values [17]. 4.4 Improvement Motivation for the redesign project The new design of LVPS bricks (v7.3) started in 2009, two years after installation, from the experience gained during operation and repairs with the idea of obtaining a general performance improvement and a solution for the problems encountered during operation. Among common problems there were spontaneous brick failures, called trips; the brick usually tripped off without damaging itself, but remained off causing data loss. During 2011 it was observed that the number of trips increased with luminosity posing serious problems to the detector functioning at higher and higher luminosity. At present, this problem persists and is constantly monitored. The graphics in Figure 4.3 shows the linear trend of the number of 25

32 trips with respect to integrated luminosity observed during Figure 4.3: This is a plot of the delivered luminosity vs the number of LVPS trips. The red points are the data and the blue line is the linear regression. The trips/pb-1 is 0.6 approximately. The data taking period used is between March 13th and September 20th By S. Norberg, J. Proudfoot, S. Chekanov. Further studies, whose results are shown in the graphics of Figure 4.4, proved that the number of trips is not a problem connected to the peak luminosity, but that depends on the integrated luminosity value. Figure 4.4: The graph shows the relationship between trip rate and peak luminosity. The red and blue points are the total number of trips divided by the integrated luminosity. The red (blue) line going through is just the average total trip rate divided by integrated luminosity for 2011 (2012). The lines show that there is very little deviation from a central value showing that there is no correlation between peak luminosity and number of trips. The rate of the trips is decreased in 2012 with resepct to 2011 beacause in this year 40 new LVPS were installed. By S. Norberg, J. Proudfoot, S. Chekanov. Since then on the solution of this problem became the major target of the redesign project. In the sections below we will see some of the improvements done to the new design and which benefits result from them Noise Improvements It was observed already during the performance studies that the noise was quite large and, as usually happens for switching power supplies, it was a function of load. Protection circuits (OVP, OCP) were on the secondary side and therefore sensitive to load current in the frontend electronics, hence they processed signals with the above mentioned noise superimposed. 26

33 In the new design OVP and OCP circuits were moved on the primary side, insulated from the load current. It was observed that the noise propagated also through ground planes, consequently in the new design the number of ground planes was increased to provide better insulation. In some areas, that still showed a lot of noise an additional filtering was added, this is the case of the output stage and also of the opto-isolator inputs and power pins. Finally the track paths were redesigned with tight loops and small areas and the feedback signals were routed with differential techniques; in particular, in the new bricks, OVP and OCP are fully differential up to the final comparator. These modifications resulted in an improvement of factor 5 on the noise level at high load currents when measured single-ended (Figure 4.5), of factor when measured with a differential probe. Figure 4.5: Single-ended output voltage measurement for +5MB bricks operating at 6A for the old brick design (left) and for the new one (right) [17] Protection of sensitive part of the circuits Some parts of the circuit such as opto-isolator and FET driver need special attention. Discussion with manufacturer led to particular protection circuits to eliminate fast spikes (in the case of LT1681) and to optimize the design of voltage connections (in the case of several components and especially the FET driver chip) Thermal Insulation In the first design thermal coupling with the cooling plate was provided by Aluminum Oxide posts placed on the bottom side of the brick, in the new design these posts have been replaced with Aluminum Nitride ones which provide a better thermal coupling. The improvement in the heat dissipation given by this arrangement can be seen from the two dimensional temperature maps of a +5MB brick operating at 6A which are shown in Figure 4.6 of the old (left) and new thermal couplings (right). Even if the obtained improvement is clear, there are two zones, the location of the converter LT1681 and the location of the FET driver, that still show a very high temperature during operation. For this reason these chips were moved from the top to the bottom side and coupled with the plate with the use of Bergquist Gap-pad [22] (a soft material with high thermal conducibility) Performance test in the calorimeter During December 2010 the first four prototype LVPS boxes were installed in the long barrel section of the tile calorimeter and the rate of voltage failures have been monitored during the following eight months of the detector operation. A plot of the number of trips for each Tile Cal module integrated over the eight months is shown in Figure 4.5. The old LVPS, shown in Figure 4.5 with the green histogram, had an average of 35 trips/module while the new ones had only one. (v7.3) to the final version (v7.5). failure for the whole period (Figure 4.7). The failure rate of a more significant number of LVPS modules obtained from production is 27

34 Figure 4.6: Two dimensional temperature maps of a +5MB brick operating at 6A with thermal coupling realized with posts of Aluminum Oxide (left) and of Aluminum Nitride (right). These maps have been obtained previous the application of Bergquist gap-pad on the LT1681 and FET driver [17]. shown in the next section. These results allowed few little adjustement on the design and Figure 4.7: The graph shows the number of LB modules that have tripped a specific number of times. This is for the period between March 13th 2011 and October 30th The red shows the new power supplies. Only one new module has tripped so far, all of the other modules have tripped on average 35 times. By S. Norberg, J. Proudfoot, S. Chekanov. encouraged a big production with the purpose of replacing all the bricks. The new version (v7.5) is at present in the final phase of production at Argonne National Laboratory (USA, IL). 28

35 4.5 Production and checkout procedure LVPS bricks are produced industrially, the components are partly procured by the industry partly from Argonne National Laboratory (ANL) where the new bricks have been designed and are being tested. ANL drafted a document addressed to the industry with the procedures and the requirements for the bricks production [23]. The paper includes precise instructions on the panel composition, soldering, cleaning and post installation. When the bricks arrive at Argonne they already have personality components and a label reporting the brick type. Each brick needs to be tested to grant the high standard of the final product shipped to CERN. The test procedure is divided in the following steps: 1. Visual inspection; 2. Initial tests (debug, repair,repetition of initial test if needed ); 3. Burn-in; 4. Final test (debug and repair if needed, repetition of final test); 5. Shipping to CERN;. First check procedure is visual inspection consisting in: checking the soldering of each part and the cleanliness of the brick and verifying the correct installation of personality components. During the visual inspection the brick receives a serial number and a barcode that will identify it. At this point the brick is logged-in in the database. The next step is the initial test, which consists of a series of tests highly automated. The test provides information about the general conditions of the brick (voltages and currents output, clock efficiency, start-up and shutdown of the brick) and checks the correct functioning of the protection and monitoring circuits. Detailed information about this test can be found in the next section If the brick fails one or more tests a technician and/or an engineer proceeds to identify the causes and to fix the problem, after that the brick will be fully retested. The initial test results, repairs and final passing date are then inserted in the database. The brick is now ready for the burn-in, a 7 hours test during which the brick runs at the upper limit current to stress components and solder joints that are on the edge of failure. During the test temperatures and input and output voltages and currents are monitored on each brick to check the correct functioning of the brick. Shut-down emergency procedures are set to prevent damages. During the burn-in bricks are thermally coupled with a water-cooled plate as it happens during standard working conditions, thermal grease is applied under the posts to help the coupling. Temperatures from the burn-in are registered with the date in database. After the burn-in the brick undergoes the final test. This test is composed by the same series of tests as the initial one. In case of need this session is followed by debug and repair; however only 2.7% of the bricks fails the final test. After repairing only the final test is repeated. The final part of the procedure is the application of the Bergquist gap-pad and the packaging for shipping. During the permanence in ANL, bricks are accompanied by a personal datasheet that contains all the informations on the tests and on the details of the failures and repairs. Whenever bricks are not under test, they are placed in electrostatic resistant bags and stored according to the type of the brick and to the test phase they are undergoing. At the end of each phase the brick receives a mark. The final product must show four marks corresponding to: visual inspection (blue), initial test passed (red), burn-in (white), final test passed (yellow). 29

36 Figure 4.8: LVPS brick ready to be shipped to CERN. Two labels are sticked on the transformer, one for the brick type and one with the serial number and barcode. At the end of each phase of testing the brick is marked with a colored dot: BLUE visual inspection, RED passed initial test, WHITE burnedin, YELLOW passed final test The test bench and the automated tests The test bench is based on a computer controlling and reading out several commercial equipments which perform the tests; the only custom component is a metal case that acts as brick support and provides the interface to the computer and the ground connections. The data are acquired primarily through a data acquisition card NI6221 [24] connected to the computer using a Peripheral Component Interconnect (PCI) interface. The data acquisition card can digitize eight channels contemporaneously and has in/out registers for control purposes. In addition there is an electronic load, a programmable high-voltage power supply, a programmable low-voltage power supply and a digital oscilloscope [25]. The full composition of the test stand can be seen in Figure 4.9. Figure 4.9: Arrangement of the test stand showing the major components and the brick [25]. In the initial and final test we check that the main functions and parameters of the bricks are correct. The required parameter are 3σ around the nominal value for all test except the over voltage and over current protections, for these tests the requirement is set by quality 30

37 assurance procedures. In the following a brief description of each test is given however the goals of these tests can be summarized as follows: ensure that the brick responds correctly to regular start-up and shut-down procedures commands and that it respects the load parameter (tests 1, 6, 13); check the voltages and the currents at the input and output stages, verify the functioning of the relative monitor and ensure that the brick responds correctly to the external reference changes (tests 2, 7, 8, 9, 10); check the output and the feedbacks sent to the LT1681, to ensure the stability of the output (test 5); ensure the correct behavior of the over voltage and over current protection circuits by forcing the brick in the condition to trigger the circuits (tests 3, 4) verify the functioning of the temperature monitors (tests 11, 12). 1. Minimum stable load This tests aims at defining the minimum load for which the power switching is behaving correctly. To determine this value the brick is connected to the nominal load value and in this condition the clock output from LT1681 is readout with the oscilloscope. The load is then decreased until missing clock cycles are registered, the load value is recorded. Permitted values for the test depends on the brick type. 2. Minimum Output Voltage This test measures the minimum output voltage. The lower limit for each brick is shown in Table 4.3, these values are within few percents of the nominal output values. Brick Type DIG+3 DIG+5 MB-5 MB+5 MB+15 HV-15 HV+5 HV+15 Minimum Output Voltage 3.3 V 5.1 V 5.1 V 5.1 V 14.5 V 14.5 V 4.8 V 14.5 V Table 4.3: Values of minimum output voltages for the different types of bricks [25]. 3. Over Voltage Protection This test checks that the over voltage protection circuitry behaves correctly. The test simulates an over voltage (for the output voltage) triggering the protection circuit that trips the brick. The condition of over voltage is reached decreasing the external reference value. The test records the voltage when the brick trips. The protection circuit is set to shutdown the brick at 10-20% over nominal operating voltage. 4. Over Current Protection This test checks that the over current protection circuitry behaves correctly. The test simulates an over current (for the output current) triggering the protection circuit that trips the brick. The condition of over current is reached decreasing the load value. The test records the current when the brick trips. The protection circuit is set to shutdown the brick at 25-50% over nominal operating current. 31

38 5. Output Analysis This test checks the stability of the output analyzing the brick output signal and the LT1681 output. The output voltage, the frequency and the duty cycle of the driving pulse are measured over several events. The maximum output voltage RMS is then required to be within the values listed in table 4.4. The driving pulse frequency must be within Âś10kHz of the nominal frequency (300kHz). The duty cycle depends on the kind of the brick and is trimmed to obtain the desired output, the RMS accepted is 0.1% for all bricks. In the table 4.4 the desired duty cycle factors for each brick are shown. The fine setting of the duty cycles helped to reduce the personality components. Brick Type Clock Maximum Duty Factor Max V out RMS DIG+3 30% 50 mv DIG+5 42% 30 mv MB-5 42% 20 mv MB+5 44% 50 mv MB+15 35% 15 mv HV-15 40% 20 mv HV+5 40% 15 mv HV+15 35% 15 mv Table 4.4: Values of clock duty cicle allowed and relative RMS and maximum V out RMS [25]. 6. Startup Verification This test measures delay at the start up of operations. For this measurement the oscilloscope probes are connected to the output and to the input of the brick to register the pulse, the delay is measured when the voltage amplitude reaches 90%. Delay acceptable values are shown in Table 4.5 Brick Type DIG+3 DIG+5 MB-5 MB+5 MB+15 HV-15 HV+5 HV+15 Start Up Delay 0.30 s ±0.10 s 0.60 s ±0.20 s 0.13 s 0± 0.08 s 0.60 s ±0.20 s 0.30 s ±0.10 s 0.12 s ±0.08 s 0.60 s ±0.20 s 0.30 s ±0.10 s Table 4.5: Values of accepted delay and relative tolerances [25]. 7. Ouput voltage versus reference voltage This test measures the response of the output at the changes of the external reference and checks the correct functioning of the output voltage monitor. The reference, sent in input to the LT1681 is slowly decreased and measurement of V out are taken at regular intervals of the reference. In the test we measure the slope and offset of V out versus the reference, then we take the same measurements for V out monitor versus offset. Knowing that the output has a well defined slope in the process, allows us to check that the output monitor sees the changing in the output and to define the proportional coefficient between the monitor and V out. The permitted values for the slopes and tolerable offsets depend on the brick type. 8. Output Current versus Load This test checks the correct functioning of the monitor circuit of output current. The brick is started at minimum output voltage, the load is 32

39 then increased up to 80% of trip point. Output of I out monitor are taken at regular intervals of the load. The test measures the slope and offset of the plot, the slope depends on the brick type. 9. Voltage Input Monitor This test checks the correct calibration of the voltage input monitor. V out is set to nominal value and the test measures the value of V in monitor. The correct value is 1 V for all bricks with a tolerance of ±0.05 V. 10. Current Input Monitor This test checks the correct calibration of the current input monitor. V out is set to the nominal value and the test measures the value of I in monitor. The correct values depends on the brick type. 11. Temperature Monitor 1 This test checks the functioning of the temperature monitor located near the input of LT1681 temperature protection. This zone is one of the hottest zones of the board. The brick is not cooled during the test therefore the measured values do not correspond to the operation values. 12. Temperature monitor 2 This test checks the functioning of the temperature monitor located near the input of LT1681 temperature protection. This zone is one of the hottest zones of the board. The brick is not cooled during the test therefore the measured values do not correspond to the operation values. 13. Group Shutdown This test checks that the brick responds correctly to various type of group shutdown commands. Inside the detector bricks are not operated in single but in groups, in various occasions the LVPS will be forced to a group shutdown. During this test the circuitry on the single brick needed for this type of shutdown is tested, with the interface board located between the test stand and the brick mimicking these instructions. The compliance with the parameter indicated ensure the correct functioning of the bricks and of the component powered. Any departure will cause severe consequences for the calorimeter module electronics. If the output (voltage or current) or the temperature of the brick exceeds the allow range it can produce damages to the brick itself and to the components it is powering. Wrong functioning of monitor circuits prevents the detection of non-critical malfunctioning therefore delaying, or even preventing, any possible response in the ELMB. A malfunctioning in the input reading and acceptance may give wrong response of the brick or prevent external intervention as for example an emergency group shutdown Summary of test results The results of the different tests for all the modules tested in the period April-September 2012 are summarized in the histograms shown in Figure 4.10 for the initial test phase (upper plot) and for the final test phase (bottom plot). Most common failures were on tests number 1, 2, 5, while tests number 3 represents a special case. Over voltage protection tuning (test #3) involves an operational amplifier, the tolerance indicated for the voltage amplification were too loose and on the 15 V brick the component-tocomponent variations with respect to the nominal value are significant for the OVP threshold setting. Almost all the 15 V bricks and a big part of the 5 V bricks required a finer tuning of the over voltage protection circuit. Usually the voltage threshold of the OVP for the 15 V bricks, as they arrived to ANL, was lower then the specifications. In some other cases the output voltage never reached the threshold thus never reached the trip conditions. On the contrary for the 5 V supplies the voltage threshold of the OVP was set over the threshold. 33

40 Figure 4.10: Statistics of failures respectively during initial test (upper) and final test (lower) for brick tested between April and September The numbers on the x axis refer to the test list in the previous section.(graphics are from ANL registers) All these problems were usually solved changing the value for the parallel R42/R112. Failures on test #1 were symptom of circuit instability and less often of real problems. In the first case repeating the test, the problem disappeared; in the second usually we operated on RA4-preload to reduce the instability at the start-up of brick operation. The output voltage (test #2) sometimes was over threshold. At one of the opto-isolator (U3) terminal there is a voltage divider, lowering the resistor on the ground branch reduces the voltage in the circuit: in many cases it was enough to solve the problem. Most of the failures of test #5 were due to the clock duty factor which was too high, these kind of failure was due to some component in the loop circuit controlling the output. In most of the cases the problematic component was the opto-isolator (U3). For the input current monitor (shown in test #10) allowed value are very low, and in many cases circuit instability was enough to give a failure, most of the time repeating the test this failure disappeared. 4.6 Production Status In August 2011 the first production of 40 LVPS box with the final design of (version 7.5) has started in ANL. In 2012, 38 of these modules were installed in the Tile Calorimeter. The number of voltage failures was monitored during the data taking period Figure 4.11 shows the number of failures for modules with the old brick design (red), with the first prototype bricks (blue) and with the final version bricks (green). The new modules did not show any failure in the period of data taking while for the old modules an average number of failures of about 90 was observed. This very good results confirmed the good design of the production version of the LVPS modules therefore the full production could be started on the remaining 260 modules. The production in ANL was on schedule during the spring and ahead of scheduled during summer. At the end of September 2012, less than 50 bricks 34