PANDA Technical Design Report - MVD

Erni W Clarkson T Cowie E Hill G Hoek M Ireland D Kaiser R Keri T Lehmann I Livingston K Lumsden S MacGregor D McKinnon B Montgomery R Murray M Protopopescu D Rosner G Seitz B and Yang G (2012) Technical Design Report for the PANDA Micro Vertex Detector Technical Report FAIR Darmstadt Copyright copy 2012 FAIR A copy can be downloaded for personal non-commercial research or study without prior permission or charge The content must not be changed in any way or reproduced in any format or medium without the formal permission of the copyright holder(s) When referring to this work full bibliographic details must be given httpeprintsglaacuk90042

Deposited on 5 February 2014

Enlighten ndash Research publications by members of the University of Glasgow httpeprintsglaacuk

Technical Design Report for the

PANDAMicro Vertex Detector

(AntiProton Annihilations at Darmstadt)

Strong Interaction Studies with Antiprotons

PANDA Collaboration





PANDA Collaboration

30th November 2011

FAIRPANDATechnical Design Report - MVD iii

The PANDA Collaboration

Universitat Basel SwitzerlandW Erni I Keshelashvili B Krusche M Steinacher

Institute of High Energy Physics Chinese Academy of Sciences Beijing ChinaY Heng Z Liu H Liu X Shen Q Wang H Xu

Universitat Bochum I Institut fur Experimentalphysik GermanyM Albrecht J Becker K Eickel F Feldbauer M Fink P Friedel FH Heinsius T Held H KochB Kopf M Leyhe C Motzko M Pelizaus J Pychy B Roth T Schroder J Schulze M Steinke

T Trifterer U Wiedner J Zhong

Universitat Bonn GermanyR Beck M Becker S Bianco K-Th Brinkmann C Hammann F Hinterberger R Jakel D Kaiser

R Kliemt K Koop C Schmidt R Schnell U Thoma P Vlasov C Wendel A WinnebeckTh Wurschig H-G Zaunick

Universita di Brescia Brescia ItalyA Bianconi

Institutul National de CampD pentru Fizica si Inginerie Nucleara ldquoHoria Hulubeirdquo Bukarest-MagureleRomania

M Bragadireanu M Caprini M Ciubancan D Pantea P-D Tarta

Dipartimento di Fisica e Astronomia dellrsquoUniversita di Catania and INFN Sezione di Catania ItalyM De Napoli F Giacoppo E Rapisarda C Sfienti

AGH University of Science and Technology Cracow PolandT Fiutowski N Idzik B Mindur D Przyborowski K Swientek

IFJ Institute of Nuclear Physics PAN Cracow PolandE Bialkowski A Budzanowski B Czech S Kliczewski A Kozela P Kulessa P Lebiedowicz

K Malgorzata K Pysz W Schafer R Siudak A Szczurek

Institute of Applied Informatics University of Technology Cracow PolandP Brandys T Czyzewski W Czyzycki M Domagala M Hawryluk G Filo D Kwiatkowski

E Lisowski F Lisowski

Instytut Fizyki Uniwersytet Jagiellonski Cracow PolandW Bardan D Gil B Kamys St Kistryn K Korcyl W Krzemien A Magiera P Moskal Z Rudy

P Salabura J Smyrski A Wronska

Gesellschaft fur Schwerionenforschung mbH Darmstadt GermanyM Al-Turany R Arora I Augustin H Deppe D Dutta H Flemming K Gotzen G Hohler

R Karabowicz D Lehmann B Lewandowski J Luhning F Maas H Orth K Peters T SaitoG Schepers CJ Schmidt L Schmitt C Schwarz J Schwiening B Voss P Wieczorek A Wilms

Veksler-Baldin Laboratory of High Energies (VBLHE) Joint Institute for Nuclear Research DubnaRussia

VM Abazov GD Alexeev VA Arefiev VI Astakhov MYu Barabanov BV BatyunyaYuI Davydov VKh Dodokhov AA Efremov AG Fedunov AA Feshchenko AS GaloyanS Grigoryan A Karmokov EK Koshurnikov VI Lobanov YuYu Lobanov AF Makarov

LV Malinina VL Malyshev GA Mustafaev AG Olshevski MA Pasyuk EA PerevalovaAA Piskun TA Pocheptsov G Pontecorvo VK Rodionov YuN Rogov RA Salmin

AG Samartsev MG Sapozhnikov GS Shabratova AN Skachkova NB Skachkov EA StrokovskyMK Suleimanov RSh Teshev VV Tokmenin VV Uzhinsky AS Vodopyanov SA Zaporozhets

NI Zhuravlev AG Zorin

iv PANDA - Strong interaction studies with antiprotons

University of Edinburgh United KingdomD Branford D Glazier D Watts P Woods

Friedrich Alexander Universitat Erlangen-Nurnberg GermanyA Britting W Eyrich A Lehmann F Uhlig

Northwestern University Evanston USAS Dobbs Z Metreveli K Seth B Tann A Tomaradze

Universita di Ferrara and INFN Sezione di Ferrara ItalyD Bettoni V Carassiti P Dalpiaz A Drago E Fioravanti I Garzia M Negrini M Savrie

G Stancari

INFN-Laboratori Nazionali di Frascati ItalyB Dulach P Gianotti C Guaraldo V Lucherini E Pace

INFN Sezione di Genova ItalyA Bersani M Macri M Marinelli RF Parodi

Justus Liebig-Universitat Gieszligen II Physikalisches Institut GermanyV Dormenev P Drexler M Duren T Eisner K Foehl A Hayrapetyan P Koch B KrıochW Kuhn S Lange Y Liang M Liu O Merle V Metag M Moritz M Nanova R Novotny

B Spruck H Stenzel C Strackbein M Thiel Q Wang

University of Glasgow United KingdomT Clarkson C Euan G Hill M Hoek D Ireland R Kaiser T Keri I Lehmann K Livingston

P Lumsden D MacGregor B McKinnon R Montgomery M Murray D Protopopescu G RosnerB Seitz G Yang

Kernfysisch Versneller Instituut University of Groningen NetherlandsM Babai AK Biegun A Glazenborg-Kluttig E Guliyev VS Jothi M Kavatsyuk P Lemmens

H Lohner J Messchendorp T Poelman H Smit JC van der Weele

Fachhochschule Sudwestfalen Iserlohn GermanyH Sohlbach

Forschungszentrum Julich Institut fur Kernphysik Julich GermanyM Buscher R Dosdall R Dzhygadlo S Esch A Gillitzer F Goldenbaum D Grunwald V Jha

G Kemmerling H Kleines A Lehrach R Maier M Mertens H Ohm DL Pohl D PrasuhnT Randriamalala J Ritman M Roeder G Sterzenbach T Stockmanns P Wintz P Wustner H Xu

University of Silesia Katowice PolandJ Kisiel

Chinese Academy of Science Institute of Modern Physics Lanzhou ChinaS Li Z Li Z Sun H Xu

Lunds Universitet Department of Physics Lund SwedenK Fissum K Hansen L Isaksson M Lundin B Schroder

Johannes Gutenberg-Universitat Institut fur Kernphysik Mainz GermanyP Achenbach A Denig M Distler M Fritsch D Kangh A Karavdina W Lauth M Michel

MC Mora Espi J Pochodzalla S Sanchez A Sanchez-Lorente C Sfienti T Weber

Research Institute for Nuclear Problems Belarus State University Minsk BelarusVI Dormenev AA Fedorov MV Korzhik OV Missevitch

Institute for Theoretical and Experimental Physics Moscow RussiaV Balanutsa V Chernetsky A Demekhin A Dolgolenko P Fedorets A Gerasimov V Goryachev

Moscow Power Engineering Institute Moscow RussiaA Boukharov O Malyshev I Marishev A Semenov

IIT Bombay Department of Physics Mumbai IndiaR Varma

FAIRPANDATechnical Design Report - MVD v

Technische Universitat Munchen GermanyB Ketzer I Konorov A Mann S Neubert S Paul M Vandenbroucke Q Zhang

Westfalische Wilhelms-Universitat Munster GermanyA Khoukaz T Rausmann A Taschner J Wessels

Budker Institute of Nuclear Physics Novosibirsk RussiaE Baldin K Kotov S Peleganchuk Yu Tikhonov

Institut de Physique Nucleaire Orsay FranceT Hennino M Imre R Kunne C Le Galliard JP Le Normand D Marchand A Maroni S OngJ Pouthas B Ramstein P Rosier M Sudol C Theneau E Tomasi-Gustafsson J Van de Wiele

T Zerguerras

Dipartimento di Fisica Nucleare e Teorica Universita di Pavia INFN Sezione di Pavia Pavia ItalyG Boca A Braghieri S Costanza A Fontana P Genova L Lavezzi P Montagna A Rotondi

Institute for High Energy Physics Protvino RussiaV Buda VV Abramov AM Davidenko AA Derevschikov YM Goncharenko VN Grishin

VA Kachanov DA Konstantinov VA Kormilitsin YA Matulenko YM Melnik AP MeschaninNG Minaev VV Mochalov DA Morozov LV Nogach SB Nurushev AV Ryazantsev

PA Semenov LF Soloviev AV Uzunian AN Vasiliev AE Yakutin

Petersburg Nuclear Physics Institute of Academy of Science Gatchina St Petersburg RussiaS Belostotski G Gavrilov A Itzotov A Kisselev P Kravchenko S Manaenkov O Miklukho

Y Naryshkin D Veretennikov V Vikhrov A Zhadanov

Kungliga Tekniska Hogskolan Stockholm SwedenT Back B Cederwall

Stockholms Universitet Stockholm SwedenC Bargholtz L Geren PE Tegner P Thoslashrngren KM von Wurtemberg

Universita del Piemonte Orientale Alessandria and INFN Sezione di Torino Torino ItalyL Fava

Universita di Torino and INFN Sezione di Torino Torino ItalyD Alberto A Amoroso MP Bussa L Busso F De Mori M Destefanis L Ferrero M Greco

T Kugathasan M Maggiora S Marcello S Sosio S Spataro

INFN Sezione di Torino Torino ItalyD Calvo S Coli P De Remigis A Filippi G Giraudo S Lusso G Mazza M Mignone A Rivetti

R Wheadon L Zotti

INAF-IFSI and INFN Sezione di Torino Torino ItalyO Morra

Politecnico di Torino and INFN Sezione di Torino Torino ItalyF Iazzi A Lavagno P Quarati K Szymanska

Universita di Trieste and INFN Sezione di Trieste Trieste ItalyR Birsa F Bradamante A Bressan A Martin

Universitat Tubingen Tubingen GermanyH Clement

The Svedberg Laboratory Uppsala SwedenB Galnander

Uppsala Universitet Institutionen for Stralningsvetenskap Uppsala SwedenH Calen K Fransson T Johansson A Kupsc P Marciniewski E Thome M Wolke J Zlomanczuk

Universitat de Valencia Dpto de Fısica Atomica Molecular y Nuclear Valencia SpainJ Dıaz A Ortiz

Warsaw University of Technology Institute of Atomic Energy Otwock-Swierk Warsaw PolandP Buda K Dmowski R Korzeniewski D Przemyslaw B Slowinski

vi PANDA - Strong interaction studies with antiprotons

Soltan Institute for Nuclear Studies Warsaw PolandS Borsuk A Chlopik Z Guzik J Kopec T Kozlowski D Melnychuk M Plominski J Szewinski

K Traczyk B Zwieglinski

Osterreichische Akademie der Wissenschaften Stefan Meyer Institut fur Subatomare Physik WienAustria

P Buhler A Gruber P Kienle J Marton E Widmann J Zmeskal

Editors K-Th Brinkmann Email Brinkmannhiskpuni-bonndeD Calvo Email calvotoinfnitT Stockmanns Email tstockmannsfz-juelichde

Technical Coordinator Lars Schmitt Email lschmittgsideDeputy Bernd Lewandowski Email blewandowskigside

Spokesperson Ulrich Wiedner Email ulrichwiednerruhr-uni-bochumdeDeputy Paola Gianotti Email paolagianottilnfinfnit

MVD Group

L Ackermann2 H Kleines6 M Ramm6

M Becker1 R Kliemt1 O Reinecke2

S Bianco1 K Koop1 J Ritman4

K-Th Brinkmann1 M Kosmata2 A Rivetti8

L Busso9 F Kruger2 R Schnell1

D Calvo8 T Kugathasan9 H Sohlbach3

S Coli8 S Marcello9 T Stockmanns4

F De Mori9 G Mazza8 J Tummo1

P De Remigis8 M Mertens4 P Vlasov1

D Deermann4 M Mignone8 R Wheadon8

S Esch4 O Morra7 Th Wurschig1

A Filippi8 J Pfennings5 P Wustner6

G Giraudo8 R Pietzsch2 H-G Zaunick1

D Grunwald5 A Pitka1 L Zotti9

R Jakel2 D-L Pohl4

V Jha4 T Quagli9

1 HISKP Universitat Bonn Germany2 IKTP TU Dresden Germany3 FH Sudwestfalen Fachbereich Informatik und Naturwissenschaften Iserlohn Germany4 Forschungszentrum Julich Institut fur Kernphysik Julich Germany5 ZAT Julich Germany6 ZEL Julich Germany7 INAF-IFSI and INFN Sezione di Torino Torino Italy8 INFN Sezione di Torino Torino Italy9 Universita di Torino and INFN Sezione di Torino Torino Italy

FAIRPANDATechnical Design Report - MVD vii

Preface

This document illustrates the technicallayout and the expected performance ofthe Micro Vertex Detector (MVD) of thePANDA experiment The MVD will detectcharged particles as close as possible tothe interaction zone Design criteria andthe optimisation process as well as thetechnical solutions chosen are discussed andthe results of this process are subjected toextensive Monte Carlo physics studies Theroute towards realisation of the detector isoutlined

viii PANDA - Strong interaction studies with antiprotons

The use of registered names trademarks etc in this publicationdoes not imply even in the absence of specific statement thatsuch names are exempt from the relevant laws and regulationsand therefore free for general use

ix

Contents

Preface vii

1 The PANDA Experiment and its Track-ing Concept 1

11 The PANDA Experiment 1

111 The Scientific Program 2

112 High Energy Storage Ring ndash HESR 4

113 Targets 6

114 Luminosity Considerations 7

12 The PANDA Detector 11

121 Target Spectrometer 11

122 Forward Spectrometer 15

123 Data Acquisition 17

124 Infrastructure 17

13 The Charged Particle Tracking System 18

131 Basic Approach 18

132 Optimisation Criteria 19

2 The Micro Vertex Detector ndash MVD 23

21 General Overview 23

22 Physics with the MVD 23

23 Basic Detector Requirements 25

24 MVD Layout 27

241 Basic Detector Geometry 27

242 Conceptual Design 28

3 Silicon Pixel Part 35

31 Hybrid Pixel Assembly Concept 35

32 Sensor 36

321 First Thinned Prototypes 36

322 Radiation Damage Study 37

323 Fullsize Prototype Sensors 39

324 Technology Choice Epi vs Oxygen 40

325 Production 41

33 Front-End Electronics 42

331 Requirements 42

332 Readout Architecture 42

333 Front-End ASIC 43

334 Module Controller 46

335 ASIC Prototypes 46

34 Hybridisation 59

341 Bump Bonding 59

35 Single Chip Assembly Prototype 60

36 Module 61

37 Bus-Flex Hybrid 62

4 Silicon Strip Part 67

41 Double-Sided Silicon Strip Detectors(DSSD) 67

411 Barrel Sensors 67

412 Wedge Sensors 70

42 Front-End Electronics ASIC 71

421 Requirements 71

422 Options 71

43 Module Data Concentrator ASIC 74

431 Architecture 74

432 Implemented Feature ExtractionAlgorithms 75

433 Implementation Status 75

44 Hybridisation 75

441 Overview 75

442 Basic Approach for the MVD 76

443 Layout of Hybrid Carrier PCB 76

444 Interconnections 78

445 Test Assembly 78

5 Infrastructure 81

51 Optical Data Transmission 81

511 GigaBit Transceiver 81

52 Off-Detector Electronics 82

53 Power Supply System 83

531 Introduction 83

532 Powering Concept for the Pixel Part 83

533 Powering Concept for the Strip Part 84

54 Cables 85

541 Requirements 85

542 Signal Cable 85

543 Power Cable 87

x PANDA - Strong interaction studies with antiprotons

55 Mechanical Structures 87

551 Global Support The Frame 87

552 The Pixel Support Structure 89

553 Support Structures of the Strip Part 92

56 The Cooling System 94

561 The Pixel Cooling System 94

562 The Strip Cooling System 98

563 Cooling Plant 99

57 DCS 101

6 Monte-Carlo Simulations and Perfor-mance 103

61 Software Framework Layout 103

62 Detector Model of the MVD 104

63 Silicon Detector Software 105

631 Monte-Carlo Particle Transport 105

632 Digitization 105

633 Noise Emulation 107

634 Local Reconstruction 107

64 Tracking and Vertexing 108

641 MVD Tracklet Finding and Fitting 108

642 Track Finding 108

643 Track Fitting 109

644 Vertex Finding Algorithms 109

65 PID Algorithm for the MVD 110

651 Energy Loss Information 110

652 Statistical Approach for the En-ergy Loss Parametrisation 110

66 General Simulation Studies 112

661 Radiation Damage 112

662 Detector Coverage 114

663 Material Budget 117

664 Rate Estimation 121

67 Resolution and Performance Studies 124

671 Hit Resolution 124

672 Vertexing Performance on Pions 124

68 Physics Channels Analysis 126

681 Detector Setup 126

682 Benchmark Channel pp rarrψ(2S)rarr Jψπ+πminus 127

683 Benchmark Channel D mesons 131

684 Considerations About the Vertexing133

7 Project Management 137

71 Quality Control and Assembly 137

711 Pixel 137

712 Strips 138

713 Integration 138

72 Safety 139

73 Timeline and Work Packages 139

Appendices 143

A The Bonn Tracking Station as a Vali-dation for the Simulation Framework 145

A1 The BonnTracking Station 145

A11 Layout of the DAQ Chain 145

A12 Structure of the Analysis Tools 145

A13 Rotation of One Sensor 146

A14 Shift of One Sensor 148

A15 Resolution Estimations 148

A16 Scattering Measurements 149

B Julich Readout System 151

B1 Overview 151

B2 Basic Concepts 151

B21 Digital Readout Board A 152

B22 Digital Readout Board B 153

B23 Firmware 153

B24 MVD Readout Framework 154

B3 Conclusion 154

C Details on Vertexing 157

D Alignment 159

D1 Introduction 159

D2 General Strategies and Techniques forVertexTracking Detectors 159

D21 Software Alignment of TrackingModules and Modern Tools 160

List of Acronyms 165

List of Figures 169

List of Tables 175

1

1 The PANDA Experiment and its TrackingConcept

The following sections contain a general introduc-tion to the PANDA experiment and in particular ashort description of the implemented overall track-ing concept They belong to a common introduc-tory part for the volumes of all individual trackingsystems

11 The PANDA Experiment

The PANDA (AntiProton ANnihilation atDArmstadt) experiment [1] is one of the keyprojects at the future Facility for Antiproton andIon Research (FAIR) [2] [3] which is currentlyunder construction at GSI Darmstadt For thisnew facility the present GSI accelerators willbe upgraded and further used as injectors Thecompleted accelerator facility will feature a com-plex structure of new accelerators and storagerings An overview of the FAIR facility is givenin figure 11 Further details of the acceleratorcomplex are described in [4] The FAIR acceleratorswill deliver primary proton and ion beams as wellas secondary beams of antiprotons or radioactiveions all with high energy high intensity and highquality Experiments to be installed at the facilitywill address a wide range of physics topics in thefields of nuclear and hadron physics as well as inatomic and plasma physics An executive summaryof the main FAIR projects can be found in [2]and [5]

The PANDA experiment will perform precise stud-ies of antiproton-proton annihilations and reactionsof antiprotons with nucleons of heavier nuclear tar-gets It will benefit from antiproton beams withunprecedented intensity and quality The coveredcentre-of-mass energy between 1 GeV and 5 GeVallows for very accurate measurements especially inthe charm quark sector Based on a broad physicsprogram studying the non-pertubative regime itwill be possible to explore the nature of the stronginteraction and to obtain a significant progressin our understanding of the QCD spectrum andhadron structure

Nowadays these studies are carried out mainly atelectron machines that offer the advantage of kine-matically clean reactions but at the price of a re-

duced set of final states and reduced cross-sectionsAlso the future experiments currently planned asupgrade at existing high-energy physics facilitieswill not deliver high-precision data over the fullcharm spectrum In this context the PANDA ex-periment will be a unique tool to improve bothstatistics and precision of existing data and to fur-ther explore the physics in the charm quark sec-tor Moreover the PANDA collaboration is in theideal situation to be able to benefit from the ex-pertise gained during the construction of the LHCdetectors and of the B-factory experiments whichhave determined a significant progress in the detec-tor technology due to the performed optimisationor the introduction of completely new concepts andinstruments

In the first section of this chapter the scientific pro-gram of PANDA will be summarised It ranges fromcharmonium spectroscopy to the search for exotichadrons and the study of nucleon structure fromthe study of in-medium modifications of hadronmasses to the physics of hypernuclei Thereforeantiproton beams in the momentum range from15 GeVc to 15 GeVc will be provided by the high-energy storage ring (HESR) to the experiment Anoverview of this accelerator and storage ring will begiven in the second section To explore the broadphysics program the PANDA collaboration wants

Figure 11 Overview of the future FAIR facility Theupgraded accelerators of the existing GSI facility willact as injectors New accelerator and storage rings arehighlighted in red experimental sites are indicated withgreen letters

2 PANDA - Strong interaction studies with antiprotons

Figure 12 Layout of the PANDA detector consisting of a Target Spectrometer surrounding the interactionregion and a Forward Spectrometer to detect particles emitted in the forward region The HESR antiprotonbeam enters the apparatus from the left side

to build a state-of-the-art general purpose detectorstudying annihilation reactions of antiprotons withprotons (pp) and in nuclear matter (pA) The differ-ent target systems will be discussed in section 113The PANDA apparatus consists of a set of systemssurrounding an internal target placed in one of thetwo straight sections of the HESR Figure 12 showsthe layout of the PANDA detector It consists of a4 m long and 2 T strong superconducting solenoidinstrumented to detect both charged and neutralparticles emitted at large and backward angles (Tar-get Spectrometer TS) and of a 2 Tm resistive dipolemagnetic spectrometer to detect charged and neu-tral particles emitted at angles between zero andtwenty degrees (Forward Spectrometer FS) withrespect to the beam axis A complex detector ar-rangement is necessary in order to reconstruct thecomplete set of final states relevant to achieve theproposed physics goals With the installed setupa good particle identification with an almost com-plete solid angle will be combined with excellentmass momentum and spatial resolution More de-tails of the PANDA detector will be described insection 12

111 The Scientific Program

One of the most challenging and fascinating goalsof modern physics is the achievement of a fullyquantitative understanding of the strong interac-tion which is the subject of hadron physics Sig-nificant progress has been achieved over the pastfew years thanks to considerable advances in exper-iment and theory New experimental results havestimulated a very intense theoretical activity and arefinement of the theoretical tools

Still there are many fundamental questions whichremain basically unanswered Phenomena such asthe confinement of quarks the existence of glueballsand hybrids the origin of the masses of hadrons inthe context of the breaking of chiral symmetry arelong-standing puzzles and represent the intellectualchallenge in our attempt to understand the natureof the strong interaction and of hadronic matter

Experimentally studies of hadron structure can beperformed with different probes such as electronspions kaons protons or antiprotons In antiproton-proton annihilation particles with gluonic degrees

FAIRPANDATechnical Design Report - MVD 3

of freedom as well as particle-antiparticle pairs arecopiously produced allowing spectroscopic studieswith very high statistics and precision Thereforeantiprotons are an excellent tool to address the openproblems

The PANDA experiment is being designed to fullyexploit the extraordinary physics potential aris-ing from the availability of high-intensity cooledantiproton beams Main experiments of the richand diversified hadron physics program are brieflyitemised in the following More details can be foundin the PANDA physics booklet [6]

bull Charmonium SpectroscopyA precise measurement of all states below andabove the open charm threshold is of funda-mental importance for a better understandingof QCD All charmonium states can be formeddirectly in pp annihilation At full luminosityPANDA will be able to collect several thousandcc states per day By means of fine scans itwill be possible to measure masses with accu-racies of the order of 100 keV and widths to10 or better The entire energy region be-low and above the open charm threshold willbe explored

bull Search for Gluonic ExcitationsOne of the main challenges of hadron physics isthe search for gluonic excitations ie hadronsin which the gluons can act as principal com-ponents These gluonic hadrons fall into twomain categories glueballs ie states of pureglue and hybrids which consist of a qq pairand excited glue The additional degrees offreedom carried by gluons allow these hybridsand glueballs to have JPC exotic quantum num-bers in this case mixing effects with nearby qqstates are excluded and this makes their exper-imental identification easier The properties ofglueballs and hybrids are determined by thelong-distance features of QCD and their studywill yield fundamental insight into the struc-ture of the QCD vacuum Antiproton-protonannihilations provide a very favourable envi-ronment in which to look for gluonic hadrons

bull Study of Hadrons in Nuclear MatterThe study of medium modifications of hadronsembedded in hadronic matter is aiming at un-derstanding the origin of hadron masses in thecontext of spontaneous chiral symmetry break-ing in QCD and its partial restoration in ahadronic environment So far experiments havebeen focussed on the light quark sector Thehigh-intensity p beam of up to 15 GeVc will al-

low an extension of this program to the charmsector both for hadrons with hidden and opencharm The in-medium masses of these statesare expected to be affected primarily by thegluon condensate

Another study which can be carried out inPANDA is the measurement of Jψ and D me-son production cross sections in p annihilationon a series of nuclear targets The compari-son of the resonant Jψ yield obtained from pannihilation on protons and different nucleartargets allows to deduce the Jψ-nucleus dis-sociation cross section a fundamental parame-ter to understand Jψ suppression in relativis-tic heavy ion collisions interpreted as a signalfor quark-gluon plasma formation

bull Open Charm SpectroscopyThe HESR running at full luminosity and atp momenta larger than 64 GeVc would pro-duce a large number of D meson pairs Thehigh yield and the well defined production kine-matics of D meson pairs would allow to carryout a significant charmed meson spectroscopyprogram which would include for example therich D and Ds meson spectra

bull Hypernuclear PhysicsHypernuclei are systems in which neutrons orprotons are replaced by hyperons In this waya new quantum number strangeness is intro-duced into the nucleus Although single anddouble Λ-hypernuclei were discovered manydecades ago only 6 double Λ-hypernuclei arepresently known The availability of p beamsat FAIR will allow efficient production of hy-pernuclei with more than one strange hadronmaking PANDA competitive with planned ded-icated facilities This will open new perspec-tives for nuclear structure spectroscopy and forstudying the forces between hyperons and nu-cleons

bull Electromagnetic ProcessesIn addition to the spectroscopic studies de-scribed above PANDA will be able to inves-tigate the structure of the nucleon using elec-tromagnetic processes such as Deeply VirtualCompton Scattering (DVCS) and the processpp rarr e+eminus which will allow the determina-tion of the electromagnetic form factors of theproton in the timelike region over an extendedq2 region Furthermore measuring the DrellYan production of muons would give access tothe transverse nucelon structure


Figure 13 Layout of the High Energy Storage Ring HESR The beam is injected from the left into the lowerstraight section The location of the PANDA target is indicated with an arrow

112 High Energy Storage Ring ndashHESR

The HESR is dedicated to supply PANDA withhigh intensity and high quality antiproton beamsover a broad momentum range from 15 GeVc to15 GeVc [7] Table 11 summarises the experimen-tal requirements and main parameters of the twooperation modes for the full FAIR version The HighLuminosity (HL) and the High Resolution (HR)mode are established to fulfil all challenging specifi-cations for the experimental program of PANDA [8]The HR mode is defined in the momentum rangefrom 15 GeVc to 9 GeVc To reach a relativemomentum spread down to the order of 10minus5 only1010 circulating particles in the ring are anticipatedThe HL mode requires an order of magnitude higherbeam intensity with reduced momentum resolutionto reach a peak luminosity of 2middot1032cmminus2sminus1 inthe full momentum range up to 15 GeVc Toreach these beam parameters a very powerful phase-space cooling is needed Therefore high-energyelectron cooling [9] and high-bandwidth stochasticcooling [10] will be utilised

The HESR lattice is designed as a racetrack shapedring with a maximum beam rigidity of 50 Tm (seefigure 13) It consists of two 180 arcs and two155 m long straight sections with a total circumfer-ence of 575 m [11] The arc quadrupole magnetswill allow for a flexible adjustment of transition en-

ergy horizontal and vertical betatron tune as wellas horizontal dispersion In the straight section op-posite to the injection point an electron cooler willbe installed The PANDA detector with the internaltarget is placed at the other side Further compo-nents in the straight PANDA section are beam in-jection kickers septa and multi-harmonic RF cavi-ties The latter allow for a compensation of energylosses due to the beam-target interaction a bunchrotation and the decelerating or accelerating of thebeam Stochastic cooling is implemented via sev-eral kickers and opposing high-sensitivity pick-upson either side of the straight sections

Figure 14 Optical functions of the γtr = 62 latticeHorizontal dispersion (a) horizontal (b) and vertical (c)betatron function Electron cooler and target are lo-cated at a length of 222 m and 509 m respectively


Experimental Requirements

Ion species Antiprotons

p production rate 2 middot 107s (12 middot 1010 per 10 min)

Momentum Kinetic energy range 15 to 15 GeVc 083 to 141 GeV

Number of particles 1010 to 1011

Betatron amplitude at IP 1 m to 15 m

Betatron amplitude E-Cooler 25 m to 200 m

Operation Modes

High resolution (HR) Peak Luminosity of 2middot1031cmminus2sminus1 for 1010 p

assuming ρtarget = 4 middot 1015 atomscm2

RMS momentum spread σpp le 4 middot 10minus5

15 to 89 GeVc

High luminosity (HL) Peak Luminosity up to 2middot1032cmminus2sminus1 for 1011 p


RMS momentum spread σpp sim 10minus4

15 to 15 GeVc

Table 11 Experimental requirements and operation modes of HESR for the full FAIR version

Special requirements for the lattice are low disper-sion in the straight sections and small betatron am-plitudes in the range between 1 m and 15 m at theinternal interaction point (IP) of the PANDA detec-tor In addition the betatron amplitude at the elec-tron cooler must be adjustable within a large rangebetween 25 m and 200 m Examples of the opti-cal functions for one of the defined optical settingsare shown in figure 14 The deflection of the spec-trometer dipole magnet of the PANDA detector willbe compensated by two dipole magnets that createa beam chicane These will be placed 46 m up-stream and 13 m downstream the PANDA IP thusdefining a boundary condition for the quadrupoleelements closest to the experiment For symmetryreasons they have to be placed at plusmn14 m with re-spect to the IP The asymmetric placement of thechicane dipoles will result in the experiment axis oc-curring at a small angle with respect to the axis ofthe straight section The PANDA solenoid will becompensated by one solenoid magnet Additionalcorrection dipoles have to be included around theelectron cooler due to the toroids that will be usedto overlap the electron beam with the antiprotonbeam Phase-space coupling induced by the elec-tron cooler solenoid will be compensated by twoadditional solenoid magnets

Closed orbit correction and local orbit bumps atdedicated locations in the ring are crucial to meetrequirements for the beam-target interaction interms of maximised ring acceptance and optimumbeam-target overlap [12] The envisaged schemeaims on a reduction of maximum closed orbit de-viations to below 5 mm while not exceeding 1 mradof corrector strength Therefore 64 beam posi-tion monitors and 48 orbit correction dipoles are

intended to be used Because a few orbit bumpswill have to be used in the straight parts of theHESR all correction dipoles therein are designed toprovide an additional deflection strength of 1 mrad

Transverse and longitudinal cooling will be usedto compensate a transverse beam blow up andto achieve a low momentum spread respectivelyWhile stochastic cooling will be applicable in thewhole momentum range electron cooling is fore-seen in a range from 15 GeVc to 89 GeVc with apossible upgrade to 15 GeVc The relative momen-tum spread can be further improved by combiningboth cooling systems Beam losses are dominatedby hadronic interactions between antiprotons andtarget protons single large-angle Coulomb scatter-ing in the target and energy straggling induced byCoulomb interactions of the antiprotons with targetelectrons Mean beam lifetimes for the HESR rangebetween 1540 s and 7100 s The given numbers cor-respond to the time after which the initial beamintensity is reduced by a factor of 1e A detaileddiscussion of the beam dynamics and beam equli-ibria for the HESR can be found in [8 13 14 15]Advanced simulations have been performed for bothcooling scenarios In case of electron cooled beamsthe RMS relative momentum spread obtained forthe HR mode ranges from 79 middot 10minus6 (15 GeVc) to27 middot10minus5 (89 GeVc) and 12 middot10minus4 (15 GeVc) [16]With stochastic cooling in a bandwidth of 2 GHz to6 GHz the RMS relative momentum spread for theHR mode results in 51 middot10minus5 (38 GeVc) 54 middot10minus5

(89 GeVc) and 39middot10minus5 (15 GeVc) [17] In the HLmode a RMS relative momentum spread of roughly10minus4 can be expected Transverse stochastic coolingcan be adjusted independently to ensure sufficientbeam-target overlap


Figure 15 Summary of the different target options foreseen at PANDA

113 Targets

The design of the solenoid magnet allows for an im-plementation of different target systems PANDAwill use both gaseous and non-gaseous targets Avery precise positioning of the target is crucial forthe exact definition of the primary interaction ver-tex In this context big challenges for either systemresult from the long distance of roughly 2 m betweenthe target injection point and the dumping systemHydrogen target systems will be used for the studyof antiproton-proton reactions A high effective tar-get density of about 4 middot 1015 hydrogen atoms persquare centimetre must be achieved to fulfill thedesign goals of the high luminosity mode Besidesthe application of hydrogen as target material anextension to heavier gases such as deuterium nitro-gen or argon is planned for complementary studieswith nuclear targets

At present two different solutions are under devel-opment a cluster-jet and a pellet target Both willpotentially provide sufficient target thickness butexhibit different properties concerning their effecton the beam quality and the definition of the IPSolid targets are foreseen for hyper-nuclear stud-ies and the study of antiproton-nucleus interactionusing heavier nuclear targets The different targetoptions are shortly described in the following Fig-ure 15 gives an overview to all target option fore-seen at PANDA

Cluster Jet Target

Cluster jet targets provide a homogeneous and ad-justable target density without any time structureOptimum beam conditions can be applied in orderto achieve highest luminosity The uncertainty ofthe IP in a plane perpendicular to the beam axisis defined by the optimised focus of the beam only

An inherent disadvantage of cluster-jet targets isthe lateral spread of the cluster jet leading to anuncertainty in the definition of the IP along thebeam axis of several millimetres

For the target production a pressurised cooled gas isinjected into vacuum through a nozzle The ejectedgas immediately condensates and forms a narrowsupersonic jet of molecule clusters The clusterbeam typically exposes a broad mass distributionwhich strongly depends on the gas input pressureand temperature In case of hydrogen the aver-age number of molecules per cluster varies from103 to 106 The cluster-jets represent a highly di-luted target and offer a very homogenous densityprofile Therefore they may be seen as a localisedand homogeneous monolayer of hydrogen atoms be-ing passed by the antiprotons once per revolutionie the antiproton beam can be focused at highestphase space density The interaction point is thusdefined transversely but has to be reconstructedlongitudinally in beam direction At a dedicatedprototype cluster target station an effective targetdensity of 15 middot1015 hydrogen atoms per square cen-timetre has been achieved using the exact PANDAgeometry [18] This value is close to the maximumnumber required by PANDA Even higher targetdensities seem to be feasible and are topic of on-going RampD work

Hydrogen Pellet Target

Pellet targets provide a stream of frozen moleculedroplets called pellets which drip with a fixed fre-quency off from a fine nozzle into vacuum The useof pellet targets gives access to high effective tar-get densities The spatial resolution of the interac-tion zone can be reduced by skimmers to a few mil-limetres A further improvement of this resolution


can be achieved by tracking the individual pelletsHowever pellet targets suffer from a non-uniformtime distribution which results in larger variationsof the instantaneous luminosity as compared to acluster-jet target The maximum achievable aver-age luminosity is very sensitive to deviations of indi-vidual pellets from the target axis The beam mustbe widened in order to warrant a beam crossingof all pellets Therefore an optimisation betweenthe maximum pellet-beam crossing time on the onehand and the beam focusing on the other is neces-sary

The design of the planned pellet target is based onthe one currently used at the WASA-at-COSY ex-periment [19] The specified design goals for thepellet size and the mean lateral spread of the pellettrain are given by a radius of 25 microm to 40 microm and alateral RMS deviation in the pellet train of approxi-mately 1 mm respectively At present typical vari-ations of the interspacing of individual pellets rangebetween 05 mm and 5 mm A new test setup withan improved performance has been constructed [20]First results have demonstrated the mono-disperseand satellite-free droplet production for cryogenicliquids of H2 N2 and Ar [21] However the pro-totype does not fully include the PANDA geometryThe handling of the pellet train over a long distancestill has to be investigated in detail The final reso-lution on the interaction point is envisaged to be inthe order of 50 microm Therefore an additional pellettracking system is planned

Other Target Options

In case of solid target materials the use of wiretargets is planned The hyper-nuclear program re-quires a separate target station in upstream posi-tion It will comprise a primary and secondary tar-get The latter must be instrumented with appro-priate detectors Therefore a re-design of the in-nermost part of the PANDA spectrometer becomesnecessary This also includes the replacement of theMVD

114 Luminosity Considerations

The luminosity L describes the flux of beam par-ticles convolved with the target opacity Hencean intense beam a highly effective target thicknessand an optimised beam-target overlap are essentialto yield a high luminosity in the experiment Theproduct of L and the total hadronic cross sectionσH delivers the interaction rate R ie the num-ber of antiproton-proton interactions in a specified

time interval which determines the achievable num-ber of events for all physics channels and allows theextraction of occupancies in different detector re-gions These are needed as input for the associatedhardware development

Obviously the achievable luminosity is directlylinked with the number of antiprotons in the HESRThe particles are injected at discrete time intervalsThe maximum luminosity thus depends on the an-tiproton production rate Rp = dNpdt Moreovera beam preparation must be performed before thetarget can be switched on It includes pre-coolingto equilibrium the ramping to the desired beammomentum and a fine-tuned focusing in the tar-get region as well as in the section for the electroncooler Therefore the operation cycle of the HESRcan be separated into two sequences related to thebeam preparation time tprep (target off) and the timefor data taking texp (target on) respectively Thebeam preparation time tprep also contains the periodbetween the target switch-off and the injection atwhich the residual antiprotons are either dumpedor transferred back to the injection momentum

Macroscopic Luminosity Profile

A schematic illustration of the luminosity profileduring one operation cycle is given in figure 16The maximum luminosity is obtained directly afterthe target is switched on During data taking theluminosity decreases due to hadronic interactionssingle Coulomb scattering and energy straggling ofthe circulating beam in the target Compared tobeam-target interaction minor contributions are re-lated to single intra-beam scattering (Touschek ef-fect) Beam losses caused by residual gas scatter-ing can be neglected if the vacuum is better than10minus9 mbar A detailed analysis of all beam loss pro-cesses can be found in [13 14] The relative beamloss rate Rloss for the total cross section σtot is givenby the expression

Rloss = τminus1 = f0 middot nt middot σtot (11)

where τ corresponds to the mean (1e) beam life-time f0 is the revolution frequency of the antipro-tons in the ring and nt is the effective target thick-ness defined as an area density given in atoms persquare centimetre For beam-target interactionsthe beam lifetime is independent of the beam in-tensity The Touschek effect depends on the beamequilibria and beam intensity At low momenta thebeam cooling scenario and the ring acceptance havelarge impact on the achievable beam lifetime


Figure 16 Time dependent macroscopic luminosity profile L(t) in one operation cycle for constant (solidred) and increasing (green dotted) target density ρtarget Different measures for beam preparation are indicatedPre-cooling is performed at 38 GeVc A maximum ramp of 25 mTs is specified for beam ac-deceleration

15 GeVc 9 GeVc 15 GeVc

Total hadronic cross section mbarn 100 57 51

Cluster jet target

Target density cmminus2 8 middot 1014 lowast 8 middot 1014 lowast 8 middot 1014 lowast

Antiproton production rate sminus1 2 middot 107 2 middot 107 2 middot 107

Beam preparation time s 120 140 290

Optimum cycle duration s 1280 2980 4750

Mean beam lifetime s sim 5920 sim 29560 sim 35550

Max Cycle Averaged Luminosity cmminus2sminus1 029 middot 1032 038 middot 1032 037 middot 1032

Pellet target

Target density cmminus2 4 middot 1015 4 middot 1015 4 middot 1015






Table 12 Calculation of the maximum achievable cycle averaged luminosity for three different beam momentaInput parameters and final results for different H2 target setups ( Lower limit cf chapter 113)

Cycle Average Luminosity

In physics terms the time-averaged cycle luminos-ity is most relevant The maximum average lumi-nosity depends on the ratio of the antiproton pro-duction rate to the loss rate and is thus inverselyproportional to the total cross section It can beincreased if the residual antiprotons after each cy-cle are transferred back to the injection momen-tum and then merged with the newly injected par-

ticles Therefore a bucket scheme utilising broad-band cavities is foreseen for beam injection and therefill procedure Basically the cycle average lumi-nosity L reads as

L = Np0 middot f0 middot nt middotτ[1minus eminus

texpτ

]texp + tprep

(12)

where Np0 corresponds to the number of availableparticles at the start of the target insertion


Target material L (pbeam=15 GeVc) L (pbeam=15 GeVc) nt

[cmminus2sminus1] [cmminus2sminus1] [atomscm2]

deuterium 5 middot 1031 19 middot 1032 36 middot 1015

argon 4 middot 1029 24 middot 1031 46 middot 1014

gold 4 middot 1028 22 middot 1030 41 middot 1013

Table 13 Expected maximum average luminosities L and required effective target thickness nt for heaviernuclear targets at PANDA at minimum and maximum beam momentum pbeam Given numbers refer to an assumednumber of 1011 antiprotons in the HESR

For the calculations machine cycles and beampreparation times have to be specified The max-imum cycle average luminosity is achieved by anoptimisation of the cycle time tcycle = texp +tprep Con-straints are given by the restricted number antipro-tons in the HESR the achievable effective targetthickness and the specified antiproton productionrate of Rp = 2 middot 107 sminus1 at FAIR

Main results of calculations performed for differ-ent hydrogen targets are summarised in table 12The total hadronic cross section σppH decreaseswith higher beam momentum from approximately100 mbarn at 15 GeVc to 50 mbarn at 15 GeVcWith the limited number of 1011 antiprotons asspecified for the high-luminosity mode cycle aver-aged luminosities of up to 16 middot1032 cmminus2sminus1 can beachieved at 15 GeVc for cycle times of less than onebeam lifetime Due to the very short beam lifetimesat lowest beam momenta more than 1011 particlescan not be provided in average As a consequencethe average luminosity drops below the envisageddesign value at around 24 GeVc to finally roughly5 middot 1031 sminus1cmminus2 at 15 GeVc Due to the lowerassumed target density the achievable luminosityof the cluster-jet target is smaller compared to thepellet operation

In case of nuclear targets the total hadronic crosssection for the interaction of antiprotons with tar-get nucleons can be estimated from geometric con-siderations taking into account the proton radiusof rp = 09 fm and the radius of a spherical nu-cleus RA which can be roughly approximated asRA = r0A

13 where r0 = 12 fm and A is the massnumber With the assumption that σppH = πr2

p the

required total hadronic cross section σpAH for a nu-cleus of mass number A can be extracted from thegiven values of σppH for antiproton-proton collisionsas follows

σpAH = π(RA + rp)2 = σppH middot

(RArp

+ 1

)2

(13)

Simulation results on maximum average luminosi-ties based on equation 13 are shown in figure 17They include adapted beam losses in the target dueto single Coulomb scattering and energy stragglingCompared to antiproton-proton experiments themaximum average luminosity for nuclear targets de-creases rapidly with both higher atomic charge Zand lower beam momenta by up to three orders ofmagnitude Specific values for selected nuclear tar-gets are given in table 13 with the effective targetthickness required to reach these numbers

Event Rates

Besides the cycle-averaged luminosity an evalua-tion of the instantaneous luminosity during the datataking is indispensable for performance studies ofthe PANDA detector Associated event rates definethe maximum data load to be handled at different

Figure 17 Maximum average luminosity vs atomiccharge Z of the target for three different beam mo-menta


Target pbeam Lexp Linst σH Rexp LpeakLexp

material [GeVc ] [cmminus2sminus1] [cmminus2sminus1] [mbarn] [sminus1] (Rnom)

hydrogen15 54 middot 1031 (59plusmn 06) middot 1031 100 54 middot 106 37

15 18 middot 1032 (20plusmn 02) middot 1032 51 97 middot 106 21

argon15 40 middot 1029 (44plusmn 04) middot 1029 2020 81 middot 105

ndash15 24 middot 1031 (26plusmn 03) middot 1031 1030 25 middot 107

gold15 40 middot 1028 (44plusmn 04) middot 1028 7670 31 middot 106


Table 14 Summary of expected event rates at PANDA Numbers for the hydrogen target correspond to thepellet system (see table 12) The given ratio LpeakLexp corresponds to the maximum value to achieve the nominalinteraction rate of Rnom = 2 middot 107 sminus1 Rough estimates for nuclear targets are based on the numbers given intable 13 with L = Lexp and σH calculated according to equation 13

timescales by the individual subsystems The dis-cussions in this section are based on the followingassumptions

bull Nominal antiproton production rate at FAIRRp = 2 middot 107 sminus1

bull Effective target densitynt = 4 middot 1015 atomscm2

bull Maximum number of antiprotons in the HESRNpmax = 1011

bull Recycling of residual antiprotons at the end ofeach cycle

As indicated in figure 16 the instantaneous lumi-nosity during the cycle changes on a macroscopictimescale One elegant way to provide constantevent rates in case of a cluster-jet target is given bythe possibility to compensate the antiproton con-sumption during an accelerator cycle by the in-crease of the effective target density Alternativelyusing a constant target beam density the beam-target overlap might be increased adequately tothe beam consumption With these modificationsthe instantaneous luminosity during the cycle is ex-pected to be kept constant to a level of 10

The values for the luminosity as given in table 12are averaged over the full cycle time Howeverto extract the luminosity during data taking Lexpthese numbers must be rescaled to consider the timeaverage over the experimental time

Lexp = (tcycletexp) middot L (14)

In addition to the fluctuation of the instantaneousluminosity during the operation cycle as dicussedabove (∆LinstLinst le 10) it must be consideredthat the HESR will be only filled by 90 in case of

using a barrier-bucket system As a consequencevalues for Linst during data taking are 10 higherthan the ones for Lexp

An estimate of peak luminosities Lpeak gt Linst mustfurther include possible effects on a short timescaleContrary to homogeneous cluster beams a distincttime structure is expected for the granular volumedensity distribution of a pellet beam Such timestructure depends on the transverse and longitudi-nal overlap between single pellets and the circulat-ing antiproton beam in the interaction region De-viations of the instantaneous luminosity on a mi-crosecond timescale are caused by variations of thepellet size the pellet trajectory and the interspac-ing between consecutive pellets The latter mustbe well controlled to avoid the possible presenceof more than one pellet in the beam at the sameinstant The resulting ratio LpeakLexp depends onthe pellet size First studies on the expected peakvalues for the PANDA pellet target have been per-formed [22] Results indicate that the peak lumi-nosity stays below 1033 cmminus2sminus1 if the pellet size isnot bigger than 20 microm

Finally for the extraction of event rates the ob-tained luminosities are multiplied with the hadroniccross section Table 14 summarises the main re-sults for a hydrogen target based on a pellet systemwhich is expected to deliver upper limits for the oc-curing event rates In addition a rough estimatefor nuclear targets based on the input of table 13and equation 13 is given Even though these valuesstill must be verified by detailed studies it can beseen that the reduced average luminosity for heaviernuclear targets is counter-balanced by an increasedcross-section that results in comparable event rates

Based on the given assumptions and caveats as dis-cussed in this section a nominal interaction rate of


Rnom = 2 middot 107 sminus1 can be defined that all detectorsystems have to be able to handle This specifica-tion includes the requirement that density fluctua-tions of the beam-target overlap have to be smallerthan a factor of two (LpeakLexp) However in orderto avoid data loss it might be important to intro-duce a generic safety factor that depends on spe-cial features of the individual detector subsystemsand their position with respect to the interactionregion

12 The PANDA Detector

The main objectives of the design of the PANDAexperiment are to achieve 4π acceptance highresolution for tracking particle identification andcalorimetry high rate capabilities and a versatilereadout and event selection To obtain a good mo-mentum resolution the detector will be composed oftwo magnetic spectrometers the Target Spectrome-ter (TS) based on a superconducting solenoid mag-net surrounding the interaction point which will beused to measure at large polar angles and the For-ward Spectrometer (FS) based on a dipole magnetfor small angle tracks An overview of the detectionconcept is shown in figure 18

It is based on a complex setup of modular sub-systems including tracking detectors (MVD STTGEM) electromagnetic calorimeters (EMC) amuon system Cherenkov detectors (DIRC andRICH) and a time-of-flight (TOF) system A so-phisticated concept for the data acquisition witha flexible trigger is planned in order to exploit at

Figure 18 Basic detection concept The main com-ponents will be described in chapter 121 and 122

best the set of final states relevant for the PANDAphysics objectives

121 Target Spectrometer

The Target Spectrometer will surround the interac-tion point and measure charged tracks in a highlyhomogeneous solenoidal field In the manner of acollider detector it will contain detectors in an onionshell like configuration Pipes for the injection oftarget material will have to cross the spectrometerperpendicular to the beam pipe

The Target Spectrometer will be arranged in threeparts the barrel covering angles between 22 and140 the forward end cap extending the anglesdown to 5 and 10 in the vertical and horizon-tal planes respectively and the backward end capcovering the region between about 145 and 170Please refer to figure 19 for an overview

Beam-Target System

The beam-target system consists of the appara-tus for the target production and the correspond-ing vacuum system for the interaction region Thebeam and target pipe cross sections inside the tar-get spectrometer are decreased to an inner diameterof 20 mm close to the interaction region The in-nermost parts are planned to be made of berylliumtitanium or a suited alloy which can be thinned towall thicknesses of 200 microm Due to the limited spaceand the constraints on the material budget close tothe IP vacuum pumps along the beam pipe canonly be placed outside the target spectrometer In-sections are foreseen in the iron yoke of the magnetwhich allow the integration of either a pellet or acluster-jet target The target material will be in-jected from the top Dumping of the target resid-uals after beam crossing is mandatory to preventbackscattering into the interaction region The en-tire vacuum system is kept variable and allows anoperation of both target types Moreover an adap-tation to non-gaseous nuclear wire targets is pos-sible For the targets of the planned hypernuclearexperiment the whole upstream end cap and partsof the inner detector geometry will be modified Adetailed discussion of the different target optionscan be found in chapter 113

Solenoid Magnet

The solenoid magnet of the TS will deliver a veryhomogeneous solenoid field of 2 T with fluctua-tions of less than plusmn2 In addition a limit of


Figure 19 Artistic side view of the Target Spectrometer (TS) of PANDA To the right of this the ForwardSpectrometer (FS) follows which is illustrated in figure 112

intBrBzdz lt 2 mm is specified for the normalised

integral of the radial field component The super-conducting coil of the magnet has a length of 28 mand an inner diameter of 90 cm using a laminatediron yoke for the flux return The cryostat for thesolenoid coils is required to have two warm boresof 100 mm diameter one above and one below thetarget position to allow for insertion of internal tar-gets The load of the integrated inner subsystemscan be picked up at defined fixation points A pre-cise description of the magnet system and detailedfield strength calculations can be found in [23]

Micro Vertex Detector

The design of the Micro Vertex Detector (MVD)for the Target Spectrometer is optimised for thedetection of secondary decay vertices from charmedand strange hadrons and for a maximum acceptanceclose to the interaction point It will also stronglyimprove the transverse momentum resolution Thesetup is depicted in figure 110

The concept of the MVD is based on radiation hardsilicon pixel detectors with fast individual pixel

readout circuits and silicon strip detectors Thelayout foresees a four layer barrel detector with aninner radius of 25 cm and an outer radius of 13 cm

Figure 110 The Micro Vertex Detector (MVD) of theTarget Spectrometer surrounding the beam and targetpipes seen from downstream To allow a look inside thedetector a three-quarters portraits is chosen


The two innermost layers will consist of pixel detec-tors and the outer two layers will be equipped withdouble-sided silicon strip detectors

Six detector wheels arranged perpendicular to thebeam will achieve the best acceptance for the for-ward part of the particle spectrum While the innerfour layers will be made entirely of pixel detectorsthe following two will be a combination of strip de-tectors on the outer radius and pixel detectors closerto the beam pipe

Additional Forward Disks

Two additional silicon disk layers are consideredfurther downstream at around 40 cm and 60 cmto achieve a better acceptance of hyperon cascadesThey are intended to be made entirely of siliconstrip detectors Even though they are not part ofthe central MVD it is planned as a first approachto follow the basic design as defined for the stripdisks of the MVD However an explicit design op-timisation still has to be performed Two of thecritical points to be checked are related to the in-creased material budget caused by these layers andthe needed routing of cables and supplies for theseadditional disks inside the very restricted space leftby the adjacent detector systems

Straw Tube Tracker (STT)

This detector will consist of aluminised Mylar tubescalled straws These will be stiffened by operatingthem at an overpressure of 1 bar which makes themself-supporting The straws are to be arranged inplanar layers which are mounted in a hexagonalshape around the MVD as shown in figure 111 Intotal there are 27 layers of which the 8 central onesare skewed to achieve an acceptable resolution of3 mm also in z (parallel to the beam) The gap tothe surrounding detectors will be filled with furtherindividual straws In total there will be 4636 strawsaround the beam pipe at radial distances between15 cm and 418 cm with an overall length of 150 cmAll straws have a diameter of 10 mm and are madeof a 27 microm thick Mylar foil Each straw tube isconstructed with a single anode wire in the centrethat is made of 20 microm thick gold plated tungstenThe gas mixture used will be Argon based with CO2

as quencher It is foreseen to have a gas gain notgreater than 105 in order to warrant long term op-eration With these parameters a resolution in xand y coordinates of less than 150 microm is expectedA thin and light space frame will hold the strawsin place the force of the wire however is kept solely

by the straw itself This overall design results in amaterial budget of 12 of one radiation length

Figure 111 Straw Tube Tracker (STT) of the TargetSpectrometer seen from upstream

Forward GEM Detectors

Particles emitted at angles below 22 which are notcovered fully by the STT will be tracked by threeplanar stations placed approximately 11 m 14 mand 19 m downstream of the target Each of thestation consists of double planes with two projec-tions per plane The stations will be equipped withGaseous micro-pattern detectors based on Gas Elec-tron Multiplier (GEM) foils as amplification stagesThe chambers have to sustain a high counting rateof particles peaked at the most forward angles dueto the relativistic boost of the reaction products aswell as due to the small angle pp elastic scatter-ing The maximum expected particle flux in thefirst chamber in the vicinity of the 5 cm diameterbeam pipe will be about 3 middot 104 cmminus2sminus1

Barrel DIRC

At polar angles between 22 and 140 particle iden-tification will be performed by the Detection of In-ternally Reflected Cherenkov (DIRC) light as re-alised in the BaBar detector [24] It will consist of17 cm thick fused silica (artificial quartz) slabs sur-rounding the beam line at a radial distance of 45 cmto 54 cm At BaBar the light was imaged across alarge stand-off volume filled with water onto 11000photomultiplier tubes At PANDA it is intendedto focus the images by lenses onto Micro-ChannelPlate PhotoMultiplier Tubes (MCP PMTs) whichare insensitive to magnet fields This fast light de-tector type allows a more compact design and thereadout of two spatial coordinates


Forward End-Cap DIRC

A similar concept is considered to be employed inthe forward direction for particles at polar anglesbetween 5 and 22 The same radiator fused sil-ica is to be employed however in shape of a diskThe radiator disk will be 2 cm thick and will havea radius of 110 cm It will be placed directly up-stream of the forward end cap calorimeter At therim around the disk the Cherenkov light will bemeasured by focusing elements In addition mea-suring the time of propagation the expected lightpattern can be distinguished in a 3-dimensional pa-rameter space Dispersion correction is achieved bythe use of alternating dichroic mirrors transmittingand reflecting different parts of the light spectrumAs photon detectors either silicon photomultipliersor microchannel plate PMTs are considered

Scintillator Tile Barrel (Time-of-Flight)

For slow particles at large polar angles particleidentification will be provided by a time-of-flight(TOF) detector positioned just outside the BarrelDIRC where it can be also used to detect photonconversions in the DIRC radiator The detector isbased on scintillator tiles of 285times 285 mm2 sizeindividually read out by two Silicon PhotoMultipli-ers per tile The full system consists of 5760 tilesin the barrel part and can be augmented also byapproximately 1000 tiles in forward direction justin front of the endcap disc DIRC Material budgetand the dimension of this system are optimised suchthat a value of less than 2 of one radiation lengthincluding readout and mechanics and less than 2 cmradial thickness will be reached respectively Theexpected time resolution of 100 ps will allow pre-cision timing of tracks for event building and fastsoftware triggers The detector also provides welltimed input with a good spatial resolution for on-line pattern recognition

Electromagnetic Calorimeters

Expected high count rates and a geometrically com-pact design of the Target Spectrometer requirea fast scintillator material with a short radiationlength and Moliere radius for the construction ofthe electromagnetic calorimeter (EMC) Lead tung-sten (PbWO4) is a high density inorganic scintilla-tor with sufficient energy and time resolution forphoton electron and hadron detection even at in-termediate energies [25 26 27]

The crystals will be 20 cm long ie approximately22 X0 in order to achieve an energy resolution be-low 2 at 1 GeV [25 26 27] at a tolerable energyloss due to longitudinal leakage of the shower Ta-pered crystals with a front size of 21times 21 cm2 willbe mounted in the barrel EMC with an inner ra-dius of 57 cm This implies 11360 crystals for thebarrel part of the calorimeter The forward end capEMC will be a planar arrangement of 3600 taperedcrystals with roughly the same dimensions as in thebarrel part and the backward end cap EMC com-prises of 592 crystals The readout of the crystalswill be accomplished by large area avalanche photodiodes in the barrel and in the backward end capvacuum photo-triodes will be used in the forwardend cap The light yield can be increased by a fac-tor of about 4 compared to room temperature bycooling the crystals down to minus25 C

The EMC will allow to achieve an eπ ratio of103 for momenta above 05 GeVc Thereforeeπ-separation will not require an additional gasCherenkov detector in favour of a very compact ge-ometry of the EMC A detailed description of thedetector system can be found in [28]

Muon Detectors

The laminated yoke of the solenoid magnet acts asa range system for the detection of muons Thereare 13 sensitive layers each 3 cm thick (layer ldquozerordquois a double-layer) They alternate with 3 cm thickiron absorber layers (first and last iron layers are6 cm thick) introducing enough material for the ab-sorption of pions in the PANDA momentum rangeand angles In the forward end cap more materialis needed due to the higher momenta of the occur-ring particles Therefore six detection layers will beplaced around five iron layers of 6 cm each withinthe downstream door of the return yoke and a re-movable muon filter with additional five layers of6 cm iron and corresponding detection layers willbe moved in the space between the solenoid andthe dipole

As sensors between the absorber layers rectangu-lar aluminum Mini Drift Tubes (MDT) are foreseenBasically these are drift tubes with additional ca-pacitive coupled strips read out on both ends toobtain the longitudinal coordinate All togetherthe laminated yoke of the solenoid magnet and theadditional muon filters will be instrumented with2600 MDTs and 700 MDTs respectively


Figure 112 Artistic side view of the Forward Spectrometer (FS) of PANDA It is preceded on the left by theTarget Spectrometer (TS) which is illustrated in figure 19

Hypernuclear Detector

The hypernuclei study will make use of the mod-ular structure of PANDA Removing the backwardend cap calorimeter and the MVD will allow to adda dedicated nuclear target station and the requiredadditional detectors for γ spectroscopy close to theentrance of PANDA While the detection of hyper-ons and low momentum Kplusmn can be ensured by theuniversal detector and its PID system a specifictarget system and a γ-detector are additional com-ponents required for the hypernuclear studies

The production of hypernuclei proceeds as a two-stage process First hyperons in particular ΞΞare produced on a nuclear target In addition asecondary target is needed for the formation of adouble hypernucleus The geometry of this sec-ondary target is determined by the short mean lifeof the Ξminus of only 0164 ns This limits the requiredthickness of the active secondary target to about25 mm to 30 mm It will consist of a compact sand-

wich structure of silicon micro-strip detectors andabsorbing material In this way the weak decaycascade of the hypernucleus can be detected in thesandwich structure

An existing germanium-array with refurbishedreadout will be used for the γ-spectroscopy ofthe nuclear decay cascades of hypernuclei Themain limitation will be the load due to neutral orcharged particles traversing the germanium detec-tors Therefore readout schemes and tracking al-gorithms are presently being developed which willenable high resolution γ-spectroscopy in an envi-ronment of high particle flux

122 Forward Spectrometer

The Forward Spectrometer (FS) will cover all parti-cles emitted in vertical and horizontal angles belowplusmn5 and plusmn10 respectively Charged particles willbe deflected by an integral dipole field Cherenkovdetectors calorimeters and muon counters ensure


the detection of all particle types Figure 112 givesan overview to the instrumentation of the FS

Dipole Magnet

A 2 Tm dipole magnet with a window frame a 1 mgap and more than 2 m aperture will be used forthe momentum analysis of charged particles in theFS In the current planning the magnet yoke willoccupy about 16 m in beam direction starting from39 m downstream of the target Thus it covers theentire angular acceptance of the TS of plusmn10 andplusmn5 in the horizontal and in the vertical directionrespectively The bending power of the dipole onthe beam line causes a deflection of the antiprotonbeam at the maximum momentum of 15 GeVc of22 For particles with lower momenta detectorswill be placed inside the yoke opening The beamdeflection will be compensated by two correctingdipole magnets placed around the PANDA detec-tion system The dipole field will be ramped duringacceleration in the HESR and the final ramp maxi-mum scales with the selected beam momentum

Forward Trackers

The deflection of particle trajectories in the field ofthe dipole magnet will be measured with three pairsof tracking drift detectors The first pair will beplaced in front the second within and the third be-hind the dipole magnet Each pair will contain twoautonomous detectors thus in total 6 independentdetectors will be mounted Each tracking detectorwill consist of four double-layers of straw tubes (seefigure 113) two with vertical wires and two withwires inclined by a few degrees The optimal angleof inclination with respect to vertical direction willbe chosen on the basis of ongoing simulations Theplanned configuration of double-layers of straws willallow to reconstruct tracks in each pair of track-ing detectors separately also in case of multi-trackevents

Forward Particle Identification

To enable the πK and Kp separation also atthe highest momenta a RICH detector is proposedThe favoured design is a dual radiator RICH de-tector similar to the one used at HERMES [29]Using two radiators silica aerogel and C4F10 gasprovides πKp separation in a broad momentumrange from 2 to 15 GeVc The two different indicesof refraction are 10304 and 100137 respectivelyThe total thickness of the detector is reduced to

Figure 113 Double layer of straw tubes with pream-plifier cards and gas manifolds mounted on rectangularsupport frame The opening in the middle of the detec-tor is foreseen for the beam pipe

the freon gas radiator (5X0) the aerogel radiator(28X0) and the aluminum window (3X0) byusing a lightweight mirror focusing the Cherenkovlight on an array of photo-tubes placed outside theactive volume

A wall of slabs made of plastic scintillator and readout on both ends by fast photo-tubes will serve astime-of-flight stop counter placed at about 7 m fromthe target Similar detectors will be placed insidethe dipole magnet opening to detect low momen-tum particles which do not exit the dipole magnetThe time resolution is expected to be in the orderof 50 ps thus allowing a good πK and Kp sepa-ration up to momenta of 28 GeVc and 47 GeVcrespectively

Forward Electromagnetic Calorimeter

For the detection of photons and electrons aShashlyk-type calorimeter with high resolution andefficiency will be employed The detection is basedon lead-scintillator sandwiches read out with wave-length shifting fibres passing through the block andcoupled to photo-multipliers The lateral size ofone module is 110 mmtimes 110 mm and a length of680 mm (= 20X0) A higher spatial resolutionwill be achieved by sub-dividing each module into 4channels of 55 mmtimes55 mm size coupled to 4 PMTsTo cover the forward acceptance 351 such modulesarranged in 13 rows and 27 columns at a distanceof 75 m from the target are required With similarmodules based on the same technique as proposedfor PANDA an energy resolution of 4

radicE [30] has

been achieved


Forward Muon Detectors

For the very forward part of the muon spectruma further range tracking system consisting of inter-leaved absorber layers and rectangular aluminiumdrift-tubes is being designed similar to the muonsystem of the TS but laid out for higher momentaThe system allows discrimination of pions frommuons detection of pion decays and with moder-ate resolution also the energy determination of neu-trons and anti-neutrons The forward muon systemwill be placed at about 9 m from the target

Luminosity Monitor

The basic concept of the luminosity monitor is to re-construct the angle of elastically scattered antipro-tons in the polar angle range from 3 mrad to 8 mradwith respect to the beam axis corresponding to theCoulomb-nuclear interference region The luminos-ity monitor will consist of a sequence of four planesof double-sided silicon strip detectors located in thespace between the downstream side of the forwardmuon system and the HESR dipole needed to redi-rect the antiproton beam out of the PANDA chicaneback into the direction of the HESR straight stretch(ie between z = + 11 m and z = + 13 m down-stream of the target) The planes are positionedas close to the beam axis as possible and are sep-arated by 10 cm along the beam direction Eachplane consists of four wafers arranged perpendic-ular surrounding to the beam axis placed at topdown right and left In this way systematic errorscan be strongly suppressed The silicon wafers willbe located inside a vacuum chamber to minimisescattering of the antiprotons before traversing thetracking planes With the proposed detector setupan absolute precision of about 3 on the time in-tegrated luminosity is considered feasible for thisdetector concept at PANDA

123 Data Acquisition

In PANDA a data acquisition concept is being de-veloped to be as much as possible matched to thecomplexity of the experiment and the diversity ofphysics objectives and the rate capability of at least2 middot 107 eventss Therefore every sub-detector sys-tem is a self-triggering entity Signals are detectedautonomously by the sub-systems and are prepro-cessed Only the physically relevant informationis extracted and transmitted This requires hit-detection noise-suppression and clusterisation atthe readout level The data related to a particle

hit with a substantially reduced rate in the prepro-cessing step is marked by a precise time stamp andbuffered for further processing The trigger selec-tion finally occurs in computing nodes which accessthe buffers via a high-bandwidth network fabricThe new concept provides a high degree of flexi-bility in the choice of trigger algorithms It makestrigger conditions available which are outside thecapabilities of the standard approach

124 Infrastructure

The experimental hall for PANDA with a plannedfloor space of 43 mtimes 29 m will be located in thestraight section at the east side of the HESR Withinthe elongated concrete cave the PANDA detectortogether with auxiliary equipment beam steeringand focusing elements will be housed Moreoverthe experimental hall will provide additional spacefor delivery of components and assembly of the de-tector parts To allow for access during HESR op-eration the beam line is shielded The floor spacefor the PANDA experiment including dipoles andfocusing elements is foreseen to have an area of37 mtimes 94 m and a height of 85 m with the beamline at a height of 35 m The general floor level ofthe HESR is planned to be 2 m higher to facilitatetransport of heavy equipment into the HESR tun-nel A crane spans the whole area with a hook at aheight of about 10 m The TS with electronics andsupplies will be mounted on rails which makes it re-tractable to parking positions outside of the HESRbeam line area In the south corner of the halla counting house complex with five floors is fore-seen It will contain supplies eg for power highvoltage cooling water and gases (1st floor) providespace for the readout electronics and data process-ing (2nd floor) as well as the online computing farm(3rd floor) and house the hall electricity supplies andventilation (5th floor) The fourth floor with thecontrol room a meeting room and social rooms forthe shift crew will be at level with the surroundingground The supply point will be at the north-eastarea of the building All other cabling which willbe routed starting at the counting house will jointhese supply lines at the end of the rails system ofthe Target Spectrometer at the eastern wall Thetemperature of the building will be moderately con-trolled More stringent requirements with respectto temperature and humidity for the detectors haveto be maintained locally To facilitate cooling andavoid condensation the Target Spectrometer willbe kept in a tent with dry air at a controlled tem-perature


Figure 114 Overview of the PANDA tracking system including the option of the additional forward disks

13 The Charged ParticleTracking System

There are different tracking systems for chargedparticles at PANDA positioned inside the targetspectrometer and in the forward region around thedipole magnet Main tasks of the global track-ing system are the accurate determination of theparticle momenta a high spatial resolution of theprimary interaction vertex and the detection ofdisplaced secondary vertices Therefore measure-ments of different subdetectors have to be mergedin order to access the full tracking information

131 Basic Approach

The magnetic solenoid field in the target spectrome-ter results in a circular transverse motion of chargedparticles with non-zero transverse momentum Theparticle momentum then can be extracted via thedetermination of the bending radius Howevertracks with a small polar angle will exit the solenoidfield too soon to be measured properly For thiscase the particle deflection induced by the subse-quent dipole magnet is used to measure the parti-cle momentum Basically it can be deduced froma combined straight line fit before and after thedipole

Due to the different analysing magnets differenttrack fitting algorithms have to be applied for cen-tral and forward tracks Central tracks are re-

constructed by combining hit points in the MVDlayers with the hit information of the STT or theGEM stations For the reconstruction of small an-gle tracks the straw tube layers in the forward spec-trometer have to be used In overlap regions theMVD the additional forward disks or the GEM sta-tions can contribute to the forward tracking becausethe delivery of an additional track point closer tothe IP significantly improves the precision of thefitting results After the global identification of in-dividual tracks an event mapping have to be per-formed to match different tracks of the same eventto a common vertex which either corresponds tothe primary interaction vertex or a delayed decayof short-lived particles

The luminosity monitor at the downstream end ofthe experiment is a tracking device of its own rightIt was introduced to measure the time integrated lu-minosity which is essential for the determination ofcross sections for different physics processes There-fore elastically scattered antiprotons are measuredunder small angles corresponding to small momen-tum transfers The associated differential cross sec-tions are well known and thus provide an ideal ref-erence channel Additional information from theMVD will eventually improve the measurement bytaking advantage of the reconstructed slow recoilproton at polar angles of around 90 which is corre-lated with the highly energetic antiproton detectedin the luminosity monitor


132 Optimisation Criteria

The different topics of the PANDA physics programwill impose specific optimisation criteria and re-quirements to design and performance of the track-ing system The optimum design thus depends onthe relative weight which is given to the differentphysics aspects Main criteria for the optimisationwill be discussed in the following

Acceptance

Full 2π azimuthal coverage is mandatory in orderto allow identification of multi-particle final statesand studies of correlations within the produced par-ticles In particular the spectroscopy program ofcharmed and strange hadrons relies on the measure-ment of Dalitz plot distributions of three-body finalstates which requires a smooth acceptance functionacross the full phase space Particular care has to betaken to avoid gaps in the acceptance function andto minimise the effect of discontinuities induced bythe transition between adjacent sub-detector com-ponents by detector frames or by mechanical sup-port structures

The fixed-target setup at PANDA implies a Lorentzboost γCM of the centre of mass ranging from 120to 292 This large dynamic range in the Lorentzboost corresponds to a large difference in the typi-cal event topologies at low and at high antiprotonmomenta At higher antiproton beam momenta thevast majority of the produced particles in the finalstate will be emitted into the forward hemisphereHowever light particles like eplusmn microplusmn or πplusmn may wellbe emitted into the backward hemisphere even athighest beam momentum As an example pionbackward emission is possible for a centre of massmomentum pcm gt 93 MeVc at pp = 15 GeVc andfor pcm gt 380 MeVc at pp = 15 GeVc

Backward charged particle tracking is needed forvarious measurements foreseen at PANDA For in-stance for the independent determination of theelectric and magnetic parts of the time-like pro-ton form factor in the reaction pp rarr e+eminus thefull angular distribution has to be measured Atq2 = 14 GeV2c2 that is at pp = 645 GeVc a po-lar angle of 160 in the centre of mass frame corre-sponds to electrons with a momentum of 070 GeVcat θlab = 113 Detection of pions in the back-ward hemisphere is important in studies of strangemulti-strange and charmed baryon resonances inpp rarr Y Y prime (+cc) reactions where the excited hy-peron Y decays by single or double pion emissionAlso higher charmonium states may emit pions with

decay energies above the critical value for backwardemission in the laboratory The PANDA trackingdetectors therefore have to cover the full range ofpolar angles between 0 and about 150

Besides the solid angle of the detector also the ac-ceptance in momentum space has to be consideredOften the final state contains charged particles withvery large and with very small transverse momen-tum components which need to be reconstructed atthe same time Given the strength of the solenoidfield of 2 T required to determine the momentumvector of the high transverse momentum particlethe radius of the transverse motion of the low trans-verse momentum particle may be small Sufficienttracking capability already at small distance fromthe beam axis is therefore mandatory As an exam-ple one may consider the reaction pp rarr Dlowast+Dlowastminus

close to threshold with Dlowast+ rarr D0π+ (amp cc) As-suming 39 MeVc momentum of the decay particlesin the Dlowastplusmn rest frame particles of the subsequentdecay D0 rarr Kminusπ+ (amp cc) have 61 MeVc mo-mentum in the D0D0 rest frame In the solenoidfield of the TS the charged pions and kaons fromthe D0D0 decay may have helix diameters up toalmost 15 m The transverse motion of the chargedpion from the Dlowastplusmn decay stays within a distance ofalmost 7 cm from the beam axis and therefore needto be reconstructed based on the track informationfrom the MVD only

Delayed Decay Vertex Detection

An important part of the PANDA physics programinvolves final states consisting of hadrons with opencharm or strangeness which decay by weak interac-tion and thus have macroscopic decay lengths Thedecay length of charmed hadrons is of the orderof 100 microm (asymp 310 microm for Dplusmn asymp 150 microm for Dplusmns asymp 120 microm for D0 asymp 130 microm for Ξ+

c asymp 60 microm forΛ+c and asymp 30 microm for Ξ0

c ) Therefore the designof the tracking system aims on a detection of de-cay vertices of particles with decay lengths above100 microm In order to achieve sufficient separationof the reconstructed decay vertex the inner part ofthe tracking system has to be located very close tothe interaction point both in longitudinal and inradial direction This requirement is fulfilled in thedesign of the MVD

The identification of hyperons and KS mesons re-quires the reconstruction of delayed decay verticesat much larger distances Λ and Ξ hyperons havecomparatively large decay lengths of about 8 cmand 5 cm respectively Due to the Lorentz boostthis may result in vertices which are displaced


by tens of centimetres from the interaction pointmostly in the downstream direction The consid-erations in the previous section concerning the re-quired acceptance thus apply with respect to theshifted emission points of charged particles Theinner part of the PANDA tracking system there-fore needs sufficient extension to the downstreamdirection in order to deliver sufficient track infor-mation for charged particle tracks originating fromthese displaced vertices

Momentum and Spatial Resolution

The spatial resolution of the tracking detectors isimportant in two aspects In the vicinity of theinteraction point it directly determines the preci-sion to which primary and displaced decay verticescan be reconstructed Further on based on the de-flection of charged particles in both solenoid anddipole magnetic fields it is an essential contributionto the momentum resolution of charged particles inall three coordinates

The detection of displaced vertices of charmedhadrons imposes particular requirements to the spa-tial resolution close the interaction point Witha typical Lorentz boost βγ 2 D meson decayvertices have a displacement of the order of a fewhundreds micrometres from the primary productionpoint Hence to distinguish charged daughter par-ticles of D mesons from prompt particles a vertexresolution of 100 microm is required The position res-olution is less demanding for the reconstruction ofstrange hadrons having decay lengths on the scaleof centimeters In this case a vertex resolution ofa few millimetres is sufficient Due to the signif-icant Lorentz boost and the small opening anglebetween the decay particles of hyperons the resolu-tion in transverse direction is required to be muchbetter than the one for the longitudinal component

The achievable momentum resolution is a complexfunction of the spatial resolution of the trackingsub-detectors the number of track-points the ma-terial budget of active and passive components re-sulting in multiple scattering the strength and ho-mogeneity of the magnetic field and of the parti-cle species its momentum and its emission angleDue to the respective momentum dependence it isgenerally expected that multiple scattering limitsthe momentum resolution of low energy particleswhereas for high energy particles the smaller curva-ture of the tracks is the dominant contribution tothe resolution

The resolution in the determination of the momen-tum vectors of the final state particles directly de-

termines the invariant or missing mass resolutionof the particles that are to be reconstructed Typi-cally the width of hadrons unstable with respect tostrong interaction (except for certain narrow stateslike eg charmonium below the DD threshold) isof the order of 10 MeVc2 to 100 MeVc2 Asan instrumental mass resolution much below thenatural width is without effect a value of a few10 MeVc2 seems to be acceptable for the identifi-cation of known states or for the mass measurementof new states With a typical scale of GeVc2 for thekinematic particle energy this translates to a rela-tive momentum resolution σpp of the order of 1as design parameter for the PANDA tracking detec-tors

Count Rate Capability

The expected count rates depend on the event rateas discussed in chapter 114 and the multiplicity ofcharged particles produced in the events While thetotal rate is of importance for DAQ design and on-line event filtering the relevant quantity for detec-tor design and performance is the rate per channelwhich is a function of the granularity per detectorlayer and of the angular distribution of the emittedparticles The latter depends on the beam momen-tum and the target material

The nominal event rate at PANDA is given by 2middot107

interactions per second In case of pp annihilationstypically only a few charged particles are producedEven if secondary particles are taken into accountthe number of charged particles per event will notbe much larger than 10 in most cases Thus thedetector must able to cope with a rate of 2 middot 108

particles per second within the full solid angle Par-ticular attention has to be paid to elastic pp scatter-ing since this process contributes significantly to theparticle load in two regions of the detector scat-tering of antiprotons at small forward angles andthe corresponding emission of recoil protons at largeangles close to 90 This affects primarily the in-ner region of the MVD disc layers and the forwardtracking detector as well as the MVD barrel partand the central tracker

The use of nuclear targets will not create signifi-cantly higher count rates than obtained with a hy-drogen or deuterium target This is due to singleCoulomb scattering which dramatically increaseswith the nuclear charge (prop Z4) and results in plosses with no related signals in the detector Incontrast to pp collisions in pA collisions no highrate of recoil particles close to 90 is expected Theemission angles of recoil protons from quasi-free pp


scattering are smeared by Fermi momentum andrescattering while recoil nuclei if they at all sur-vive the momentum transfer are too low energeticto pass through the beam pipe

Particle Identification

Charged particle identification over a wide rangeof momentum and emission angle is an essentialprerequisite for the capability of PANDA to accom-plish the envisaged physics program Charged par-ticles with higher momenta will be identified viaCherenkov radiation by the DIRC detector in theTarget Spectrometer and by the forward RICH de-tector in the Forward Spectrometer For positivecharged kaon-pion separation in the DIRC about800 MeVc momentum is required While almost allparticles emitted within the acceptance of the For-ward Spectrometer are above the Cherenkov thresh-old due to the forward Lorentz boost a number ofinteresting reaction channels have final states withheavier charged particles (Kplusmn p p) at larger angleswith momenta below the DIRC threshold In orderto separate these low energy kaons from the muchmore abundant pions particle identification capa-bility based on energy loss information has to besupplied by the central tracking detector

Material Budget

Any active or passive material inside the detectorvolume contributes to multiple scattering of chargedparticles electron bremsstrahlung and photon con-version and thus reduces the momentum resolutionfor charged particles in the tracking detectors anddetection efficiency and energy resolution for pho-tons in the EMC Therefore the material budgethas to be kept as low as possible Following themore demanding requirements to meet the perfor-mance criteria of the EMC a total material budgetof MVD and Central Tracker below 10 is still con-sidered to be acceptable [28]

References

[1] PANDA collaboration Letter of intent forPANDA - Strong Interaction Studies with An-tiprotons Technical report FAIR-ESAC 2004

[2] GSI Helmholtzzentrum fur Schwerionen-forschung FAIR - An International Acceler-ator Facility for Beams of Ions and Antipro-tons Baseline technical report 2006 http

wwwgsidefairreportsbtrhtml

[3] FAIR - Facility for Antiproton and IonResearch Green Paper - The Modular-ized Start Version Technical Report2009 httpwwwfair-centerde

fileadminfairpublications_FAIR

FAIR_GreenPaper_2009pdf

[4] P Spiller and G Franchetti The FAIR ac-celerator project at GSI Nucl Instr MethA561305ndash309 2006

[5] WF Henning FAIR - recent developmentsand status Nucl Instr Meth A805502cndash510c 2008

[6] PANDA collaboration Physics PerformanceReport for PANDA - Strong Interaction Stud-ies with Antiprotons arXiv09033905v1 [hep-ex] 2009

[7] FAIR Technical Design Report HESRTechnical report Gesellschaft fur Schwerio-nenforschung (GSI) Darmstadt April 2008httpwww-wingsideFAIR-EOIPDF

TDR_PDFTDR_HESR-TRV312pdf

[8] A Lehrach et al Beam Dynamics of the High-Energy Storage Ring (HESR) for FAIR IntJ Mod Phys E 18(2)420ndash429 2009

[9] B Galnander et al Status of ElectronCooler Design for HESR Proc of theEuropean Accelerator Conference EPAC2008GenoaProc of the European Accelerator Con-ference EPAC2008 Genoa THPP(049)3473ndash3475 2008

[10] H Stockhorst D Prasuhn R Maier and BLorentz Cooling Scenario for the HESR Com-plex AIP Conf Proc 821190ndash195 2006

[11] B Lorentz et al HESR Linear Lattice DesignProceedings of European Accelerator Confer-ence EPAC Genova MOPPC(112)325ndash3272008

[12] DM Welsch et al Closed Orbit Correc-tion and Sextupole Compensation Schemes forNormal-Conducting HESR In the EuropeanAccelerator Conference EPAC2008 GenoaTHPC076 p 3161 2008

[13] A Lehrach et al Beam Performance and Lumi-nosity Limitations in the High-Energy StorageRing (HESR) Nucl Instr Meth A561289ndash296 2006

[14] F Hinterberger Beam-target Interaction andIntrabeam Scattering in the HESR TechnicalReport FZJ JUL-4206 2006 ISSN 0944-2952


[15] O Boine-Frankenheim R Hasse F Hinter-berger A Lehrach and P Zenkevich Cool-ing Equilibrium and Beam Loss with InternalTargets in High Energy Storage Rings NuclInstr Meth A560245ndash255 2006

[16] D Reistad et al Calculations on High-EnergyElectron Cooling in the HESR In Proceed-ings of the Workshop on Beam Cooling andRelated Topics COOL2007 Bad KreuznachMOA2C05 p 44 2007

[17] H Stockhorst et al Stochastic Cooling Devel-opments for the HESR at FAIR Proc of theEuropean Accelerator Conference EPAC2008GenoaProc of the European Accelerator Con-ference EPAC2008 Genoa THPP(055)3491ndash3493 2008

[18] A Taschner E Kohler H-W Ortjohann andA Khoukaz High Density Cluster Jet Tar-get for Storage Ring Experiments Nucl InstrMeth A660(1)22ndash30 2011

[19] Chr Bargholtz et al Properties of the WASApellet target and a stored intermediate-energybeam Nucl Instr Meth A587178ndash187 2008

[20] M Buscher et al The Moscow-Julich Frozen-Pellet Target AIP Conf Proc 814614ndash6202006

[21] M Buscher et al Production of HydrogenNitrogen and Argon Pellets with the Moscow-Julich Pellet Target Int J Mod Phys E18(2)505ndash520 2009

[22] A Smirnov et al Effective luminosity simula-tion for PANDA experiment at FAIR In Pro-ceedings of COOL2009 Lanzhou China TH-PMCP002 2009

[23] PANDA collaboration Technical Design Re-port for the Solenoid and Dipole SpectrometerMagnets arxiv09070169v1 [physicsins-det]2009

[24] H Staengle et al Test of a large scale proto-type of the DIRC a Cherenkov imaging detec-tor based on total internal reflection for BaBarat PEP-II Nucl Instr Meth A397261ndash2821997

[25] K Mengel et al Detection of monochromaticphotons betwen 50 MeV and 790 MeV with aPbWO-4 scintillator array IEEE Trans NuclSci 45681ndash685 1998

[26] R Novotny et al Electromagnetic calorimetrywith PbWO-4 in the energy regime below 1GeV IEEE Trans Nucl Sci 471499ndash15022000

[27] M Hoek et al Charged particle response ofPbWO-4 Nucl Instr Meth A486136ndash1402002

[28] PANDA collaboration EMC Technical DesignReport Technical report Darmstadt 2008arXiv08101216v1

[29] N Akopov et al The HERMES dual-radiatorring imaging Cerenkov detector Nucl InstrMeth A479511ndash530 2002

[30] I-H Chiang et al KOPIO - a search for K0 rarrπ0νν 1999 KOPIO Proposal

23

2 The Micro Vertex Detector ndash MVD

21 General Overview

This volume illustrates the technical layout and theexpected performance of the Micro Vertex Detector(MVD) of the PANDA experiment The introduc-tory part contains a physics motivation that un-derlines the importance of this tracking system forkey experiments at PANDA Basic detector require-ments are defined taking into account both the de-mands of the scientific program and the experimen-tal conditions of the antiproton accelerator HESRas well as the PANDA target and detector setupMoreover the overall layout of the MVD is de-scribed

Silicon detectors will be used as sensitive elementsinside the MVD The default option foresees hybridpixel detectors and double-sided microstrip detec-tors in the inner and outer parts respectively Bothof them are based on different detector technologiesand thus impose specific requirements for the sen-sor production and characterisation the connectedfront-end electronics and the hybridisation Henceall these topics will be discussed separately for thesilicon pixel and the silicon strip part in chapter 3and 4 respectively Further aspects related to theMVD infrastructure which are relevant for bothparts will be outlined in chapter 5 The perfor-mance of the MVD has been studied in extensiveMonte Carlo simulations which will be summarisedin chapter 6

22 Physics with the MVD

Charged particle tracking is one of the essentialparts in the overall detection concept at PANDAwhich aims at fully exclusive measurements with aflexible trigger The MVD delivers 3D hit infor-mation close to the interaction point which in ad-dition defines the earliest possible measurement ofparticles in the reaction channel Track and timeinformation both provide important input for thesubsequent event reconstruction Precise timing isessential for an accurate assignment of individualtracks to the appropriate event In addition afast detection of particles inside the MVD can beused as time reference for outer detector systemsThe track information in the magnetic field of thesolenoid gives access to the particle momentum andthe event topology

The identification of open charm and strangenessis one of the major tasks for the experiment andhinges on the capability to resolve secondary decaysof short-lived particles in displaced vertices otherthan the primary reaction point In addition theenergy loss of slower protons kaons and pions inthe silicon detectors may be used as well to con-tribute to the global particle identification Someexamples of decay channels that hold the potentialto be identified by purely geometrical means arelisted in table 21 Basically strange and charmedhadrons exhibit two different decay length scales inthe order of several centimetres and a few hundredmicrometres respectively While PANDA will fo-cus on charm production the identification of kaonswill greatly enhance the efficiency also for D mesonssince they show large branchings into channels ac-companied by kaons

Key points of the PANDA physics program pre-sented in chapter 111 can only be met with a suit-able vertex resolution In this context the MVDwill be of central importance for the associatedphysics goals An identification of ground state Dmesons via their displaced vertex is of particular in-terest because they can be produced in associateddecays of excited D mesons as well as charmoniaor charmonium-like states above the DD thresh-old The applicability of a precise D meson taggingis thus essential for the spectroscopy in the opencharm sector and the charmonium mass region

In addition the experimental program at PANDAputs emphasis on the high-resolution spectroscopyof electrons Electrons are excellent tags for D me-son spectroscopy (cf table 21) They carry in-formation on the flavour of the decaying charmWith heavy targets leptons can also be used toreconstruct mesonic properties inside the nuclearmedium

Some of the envisaged physics benchmark channelswhich all require the use of the vertex detector forquite different reasons are listed in table 22 Themost stringent requirements are given by those re-action channels where the D decay will be used toselect certain event patterns from the data streamduring data acquisition as indicated in the last col-umn of the table


particle lifetime decay length cτ decay channel (fraction)

K0S 8953(5) ps 26842 cm π+πminus ((6920plusmn 005))

Dplusmn 1040(7) ps 3118 microm e+semileptonic + cc ((1607plusmn 030))

Kminusanything + cc ((257plusmn 14))

K+anything + cc ((59plusmn 08))

K0anything +K0anything ((61plusmn 5))

eg Kminusπ+π+ + cc ((94plusmn 04))

K0π+π+πminus + cc ((68plusmn 029))

D0 4101(15) fs 1229 microm e+anything + cc ((649plusmn 011))

micro+anything + cc ((67plusmn 06))




eg K0SK

+Kminus ((465plusmn 030))

Kminusπ+π+πminus + cc ((809+021minus018))

K0Sπ

+πminusπ0 + cc ((54plusmn 06))

Dplusmns 500(7) fs 1499 microm e+semileptoniccc ((65plusmn 04))



eg K+Kminusπ+ + cc ((550plusmn 027))

Λ 2631(20) ps 789 cm pπminus ((639plusmn 05))

Σ+ 8018(26) ps 2404 cm pπ0 ((5157plusmn 030))

nπ+ ((4831plusmn 030))

Σminus 1479(11) ps 4434 cm nπminus ((99848plusmn 0005))

Ξminus 1639(15) ps 491 cm Λπminus ((99887plusmn 0035))

Ωminus 821(11) ps 2461 cm ΛKminus ((678plusmn 07))

Ξ0πminus ((236plusmn 07))

Ξminusπ0 ((86plusmn 04))

Λ+c 200(6) fs 599 microm pK

0((23plusmn 06)

pKminusπ+ ((50plusmn 13)

Λπ+π+πminus ((26plusmn 07)

Σ+π+πminus ((36plusmn 10)

Ξ0c 112(13) fs 336 microm Ξminusπ+(not known)

Table 21 Strange and charmed candidates for identification by means of their delayed decay as listed in [1]

reaction channel detected particle tracking used for

pp rarr φφ 2K+ 2Kminus momentum measurement (PID)

pp rarr ηc Kplusmn π∓ K0S momentum measurement (PID)

pp rarrDD Krsquos and πrsquoscharm detection online

momentum measurement (PID)

pA rarrDD X D (D) inclusive charm ID online

pp rarr ψ(2S) π+ πminus Jψ (rarr e+eminus micro+microminus) vertex constraint

Table 22 Selection of benchmark channels requiring an optimum performance of the vertex tracking The mostserious challenge will be D identification due to the very short decay lengths of these mesons


Figure 21 Expected particle distribution inantiproton-proton collisions obtained with 107 DPM [2]events at different beam momenta (pbeam) Top Particledistributions as a function of particle momentum andpolar angle The red line visible in both plots resultsfrom the elastic scattering process that is implementedin the DPM generator Bottom 1D profile projectedonto the polar angle

23 Basic DetectorRequirements

Basic detector requirements are mandated by thephysics goals of the PANDA experiment and the ex-perimental conditions the detector will meet duringoperation The MVD has to cope with differentexperimental setups of the target region In anycase before reaching the MVD particles must tra-verse at least one wall of the beam or target pipeThe minimum particle momentum for different par-ticle species needed to reach the inner layer of theMVD is listed in table 23 Figure 21 illustrates thedistribution of final state particles in antiproton-proton collisions at different beam momenta whichis expected to be obtained with a hydrogen targetintended to be used in the first stage of PANDA

The fixed-target setup is reflected by a Lorentzboost of particles in the forward direction whichincreases with higher beam momentum Most ofthese particles carry away a significant part of theinitial beam momentum and thus the momentum

Figure 22 Expected particle distribution inantiproton-nucleus collisions obtained at maximumbeam momenta (pbeam) Top Particle distributions asa function of particle momentum and polar angle ob-tained with 40000 UrQMD [3] events Bottom 1Dprofile projected onto the polar angle obtained with ahigher statistical sample of 2times 106

Particle |~pmin| αinc = 90 |~pmin| αinc = 45

p p sim89 MeVc sim98 MeVc

πplusmn sim23 MeVc sim25 MeVc

Kplusmn sim56 MeVc sim62 MeVc

microplusmn sim19 MeVc sim21 MeVc

eplusmn sim1 MeVc sim1 MeVc

Table 23 Minimum momentum |~pmin| for different par-ticle species needed to traverse the 200 microm thick beryl-lium pipe intended to be used at the beam-target regionassuming two diffrent incident angles αinc Calculationsare based on the Bethe-Bloch energy loss

range spans more than two orders of magnitudeSlower particles in a range of 100 MeVc up to500 MeVc are emitted more or less isotropic overnearly the full solid angle The elastic scatteringprocess in antiproton-proton reactions delivers anenhanced emission of slow recoil protons at polarangles above 80 Results on the particle distribu-tion in the final state to be expected with heav-ier targets are shown in figure 22 In comparisonwith antiproton-proton collisions the Lorentz boostin forward direction in antiproton-nucleus reactions


is less distinct and vanishes with increasing atomicnumber

Central task of the MVD is a precise measurementof both the primary interaction vertex and sec-ondary decay vertices of short-lived particles Forthis purpose the detection of a first track point veryclose to the nominal vertex is mandatory Moreoverthe obtained track information from the vertex de-tector also results in an improved momentum res-olution The general detection concept of PANDArequests coverage over nearly the full solid angleThe boundary conditions of adjacent detector sub-systems define a range of approximately 3 to 150

in polar angle to be covered by the MVD

Due to the fixed-target setup a very good spatialresolution is particularly needed in forward direc-tion Therefore a high granularity and intrinsic de-tector resolution are a prerequisite Slightly lowerquality requirements result for the backward hemi-sphere A sufficient number of track points must bemeasured inside the MVD to achieve an optimumtracking performance In the best case an indepen-dent pre-fit of MVD tracklets can be accomplishedfor which a minimum of four track points is neces-sary Further improvements can be achieved by ananalogue readout of the signal amplitudes It facili-tates a cluster reconstruction over detector elementssharing the signal and additional energy loss mea-surements which would be of advantage as theycould be used in the particle identification at lowenergy

Another main issue for the MVD is the minimisa-tion of the material budget PANDA will have tocope with a large amount of low-energy particlesScattering in the silicon layers is thus a major con-cern since the small-angle scattering drastically in-creases with decreasing particle energy Moreoverany material near the primary target may causebackground due to bremsstrahlung and pair pro-duction The loss of photons related to the conver-sion or absorption in the detector material signifi-cantly affects the overall performance of the PANDAapparatus

As a consequence of the close position with re-spect to the interaction region all detector compo-nents must withstand an adequate radiation doseWith an overall lifetime of PANDA of about 10years and an assumed duty cycle of 50 therequested radiation tolerance given in terms of1 MeV neutron equivalent particles per squarecentimetre 1 n1MeV eq cmminus2 is in the order of1014 n1MeV eq cmminus2 [4] and thus calls for dedicatedproduction techniques The achievable time resolu-tion of the MVD must be in the order of 10 ns in

order to cope with the high interaction rate Par-ticular challenges arise from the specific trigger andreadout concept of the PANDA experiment that isbased on an autonomous readout scheme for eachof the individual detector sub-systems Thereforean internal trigger and a first data concentrationmust be integrated for each channel and on front-end level respectively This approach exceeds com-mon state-of-the-art solutions and required an ex-tended research and development program

In the following the most important detector re-quirements will be listed to conclude this section

bull Optimum detector coverageWithin the given boundary conditions a full de-tector coverage must be envisaged The num-ber of layers is defined by the design goalwhich foresees a minimum of four MVD hitpoints per track An increased number of lay-ers is necessary in the forward part in order tofulfill the required tracking performance

bull High spatial resolutionTracks of charged hadrons have to be measuredwith a spatial resolution no worse than 100 micromin z and a few tens of microm in xy The achievablevertex resolution shall be in the same rangeof 100 microm in order to resolve displaced decayvertices of open-charm states such as the D0

and the Dplusmn with mean decay lengths of 123 micromand 312 microm respectively (cf table 21) Oneof the most demanding tasks is the recognitionof DD events as decay products of charmoniumstates via their displaced vertices At energiesnot much above the threshold these mesonswill be strongly forward focused In this casethe best resolution is required along the beam(z-direction)

bull Low material budgetThe impact of the MVD on the outer detec-tor systems must be kept as low as possi-ble Photon conversion in particular close tothe interaction point has to be minimised forthe efficient operation of the electromagneticcalorimeter Moreover scattering deterioratesthe overall tracking performance In view ofboth deteriorating effects the total materialbudget in units of radiation length XX0 shallstay below a value of XX0 = 10 Besidesthe sole reduction of the total material amountlightweight and low-Z materials must be usedwhere ever possible

bull Improved momentum resolutionThe track information of the MVD defines im-


portant constraints for the applied tracking al-gorithms and thus facilitates the determinationof the particle momentum In this context itis of special importance for small-angle tracksWith the minimised material budget mattereffects will be reduced to such a level that theMVD can improve the overall momentum res-olution by roughly a factor of two

bull Additional input to the global particleidentificationThe lower momenta for most of the emittedparticles facilitate an improved particle iden-tification based on energy loss measurementswhich can provide additional to sustain differ-ent particles hypothesis

bull Fast and flexible readoutThe readout has to be designed in a differentway as compared to other experiments in or-der to allow continuous data collection fromthe detector without external triggering Asmultiplicities are expected to be rather small(a maximum of 16 charged particles was esti-mated for the studies envisaged) the more se-rious aspect of data extraction from the vertexdetector is caused by the rather high interac-tion rates which will be in excess of 107 an-nihilations per second The occupancy in dif-ferent detector regions is very anisotropic andwill also change due the use of different targetsThis requires an efficient and flexible readoutconcept

24 MVD Layout

The MVD is the innermost part of the central track-ing system Its position inside the PANDA targetspectrometer the beam-target geometry and the re-sulting boundary conditions are shown in figure 23In the following subsections the basic geometry andthe conceptual design of the detector will be dis-cussed

241 Basic Detector Geometry

The MVD is divided into a barrel and a forwardpart The outer limits are given by the detectorposition inside the central tracker which defines amaximum radius of 15 cm The extension of theMVD along the beam axis is roughly z = plusmn23 cmwith respect to the nominal interaction point Aschematic picture of the MVD layers is shown in fig-ure 24 There are four barrel layers and six forward

disks Silicon hybrid pixel detectors are foreseen inthe two innermost barrel layers and for all disksDouble-sided silicon micro-strip detectors will beused in the two outer barrel layers and as radialcomplements in the last two disk layers The basicdetector layout results in a detector coverage witha minimum of four track points in a polar angleinterval from 9 to 145

The barrel part of the MVD covers polar angles be-tween 40 and 150 The maximum downstreamextension is given by both outer strip layers Thetwo pixel layers end at polar angles around 40

thus avoiding shallow crossing angles of particlesThe upstream extension of all barrel layers is cho-sen to fit with the opening cone of the beam pipeThe radii of the innermost and the outermost bar-rel layer are set to 25 cm and 135 cm respectivelyThe two intermediate barrel layers are arranged inincreasing order

Figure 23 Position of the MVD inside the targetspectrometer (a) and zoom of the beam-target geome-try (b) The MVD volume is highlighted in green theassociated readout in upstream direction is indicatedby green arrows Boundary conditions given by adja-cent sub-systems ie the beam and target pipes as wellas the outer tracker are visualised by red arrows Allmeasures refer to a maximum extension of the MVDvolume without any safety margins


Figure 24 Basic layout of the MVD (top) The red in-ner parts are equipped with silicon hybrid pixel sensorsDouble-sided silicon micro-strip detectors utilised in theouter layers are indicated in green Bottom Sideviewalong the beam axis illustrating the polar angle cover-age The barrel and the forward part meet at a polarangle of θ = 40

The disk layers in forward direction enable a mea-surement at small polar angles between 3 and 40The innermost disk located at z = 2 cm is the clos-est of all detector layers with respect to the nominalinteraction point It has an interspacing of 2 cm tothe second pixel disk Both of these small pixeldisks are located inside the outer pixel barrel layerFurther downstream there are four large pixel disksWhile the first two of them are positioned inside thestrip barrel layers the ones still further downstreamare outside the barrel layers and extended to largerradii by additional strip disks

In addition to the MVD there are two extra disksenvisaged in forward direction that would fill thelong detector-free gap to the forward GEM track-ing station They are intended to contribute to thevertex reconstruction of hyperons which have muchlonger lifetimes and consequently a larger displace-ment of the secondary vertex than D mesons Theconceptual design of these additional disks is similarto that of the strip disks of the MVD

242 Conceptual Design

Choice of Detector Technology

Silicon detectors excel in a fast response and a lowmaterial budget Moreover they allow a high de-gree of miniaturisation and they can be produced inbig quantities with very good reproducibility Dueto these properties they perfectly meet the require-ments imposed on the MVD

Silicon hybrid pixel detectors deliver discrete 2D in-formation with a high granularity and allow veryprecise space point measurements They are in-tended to be used in the innermost MVD layers inorder to cope with the high occupancy close to theinteraction region However the total number ofchannels required to cover larger surfaces increasesrapidly

The material budget of the detector scales with thenumber of readout channels It must be minimisedin order to fulfill the requirements of the PANDAexperiment For this reason double-sided siliconmicro-strip detectors are foreseen in the outer partsof the MVD They facilitate the readout of a muchlarger area with significantly fewer channels How-ever an utilisation very close to the primary vertexis disfavored because of the high probability of mul-tiple hits in the detector which lead to ambiguitiesso-called ghost hits and the increased probabilityfor hit loss due to pile-up

In the following all individual silicon detectors willbe denoted as either pixel or strip sensors in orderto avoid misinterpretations

Hierarchical Structure

The hierarchical structure of the MVD is based ona modular concept following the future test and in-stallation sequence of the detector The silicon sen-sors represent the lowest level therein A detectormodule is defined as the smallest functional unitwhich is electronically independent It is formed byall hard-wired connections between individual sen-sors and assigned components of the readout elec-tronics

The finalised hybrid is formed by a super-module in-cluding several detector modules and the associatedcooling and support structure It corresponds to thesmallest mechanically independent unit for the de-tector assembly Different detector layers are thencomposed of super-modules which are attached tothe respective mechanical holding structure In thisway four main building blocks are created Theyare given by the individual pixel and strip layers


Figure 25 Hierarchical structure of the MVD

which form two half-shells in the barrel part andtwo half-disk structures in the forward part respec-tively The half-detector concept is required due tothe target pipe which crosses the entire MVD vol-ume from the top to the bottom and thus breaksrotational symmetry around the beam axis As aconsequence the detector will be left-right sepa-rated along its middle plane A schematic pictureof the hierarchy is given in figure 25

Sensor Geometry

The schematic detector layers as shown in figure 24must be approximated by an appropriate arrange-ment of individual detector elements The precisesensor dimensions and the exact sensor positioningresult from an explicit design optimisation [5] Ithas been accomplished to facilitate compromise be-tween good spatial coverage and minimum materialon the one hand and the introduction of sufficientspace needed for passive elements and appropriatesafety margins for the detector integration on theother Therefore technical aspects have been takeninto account alongside with purely geometric con-siderations The final number of different sensortypes is kept small in order to facilitate the modu-lar concept and to obtain a high compatibility Thespecific size of the readout structure has direct im-pact on the sensor design For the two differentdetector types it is given by the pixel cell size andthe strip interspacing respectively

In case of pixel sensors a rectangular shape is fa-vored due to technical reasons The PANDA pixel

Figure 26 Schematics of the basic pixel sensor geom-etry Top Main pixel sensor types The design is basedon a quadratic pixel cell size of 100 times 100 microm2 and aneffective readout area of approximately 1 cm2 per indi-vidual readout chip Bottom Pixel sensor arrangementin different detector layers

design is based on a quadratic pixel cell size witha side length of 100 microm There are four differentpixel sensors each of them has the same width butthey differ in lengths The dimensions correspondto multiples of the readout matrix of the connectedreadout chip which serves 116times110 cells (see chap-ter 33) and thus features an area of approximately13 cm2 The different pixel sensor types are shownschematically in figure 26 on the top


Figure 27 Basic sensor geometry for the MVD strip part Sensor types (top) and sensor arrangement in thedifferent detector layers (bottom) The underlying segmentation of the sensors is given by the stereo angle andthe strip pitch assuming 128 channels per readout chip

The basic geometry for the pixel sensor arrange-ment is illustrated below A pixel barrel layeris made of a double-ring arrangement in order toachieve sufficient radial overlap Shorter sensors areused to obtain sufficient space for the target pipelead-through The forward disk layers consist of anappropriate configuration of pixel sensors which arealigned in different rows Alternating rows are po-sitioned with a small interspacing along the beamaxis thus allowing the compensation of the passiveedge around the active area

For silicon strip detectors there is no direct impactof the readout chip size on the dimension of thesensor segmentation Hence the distance betweenneighbouring strips which is commonly denoted asldquostrip pitchrdquo is one of the main parameters forthe detector optimisation An approximate rangeis given by the average track deviation related toscattering effects in the previous pixel layers Cal-culated values for the mean scattering angle andthe associated deflection of tracks are shown in fig-ure 28 They are obtained with an empirical for-mula for the multiple scattering in a Gaussian ap-proximation given by Lynch and Dahl [6]

Figure 28 Mean scattering angle σ(Θ) and the as-sociated particle deflection σperp as a function of the par-ticle momentum p for two effective silicon layer thick-nesses Values for electrons pions and protons are cal-culated with a Gaussian approximation [6] The rele-vant momentum range for the strip barrel layer of theMVD is highlighted in yellow


Basic parameter Pixel part Strip part

Number of super-modules 66 70

Number of detector modules 176 140

Number of sensors

34 (2 chips size) 172 (rectangular)

28 (4 chips size) 34 (squared)

54 (5 chips size) 48 (trapezoidal)

60 (6 chips size)

Total 176 254

Active silicon area [m2] 0106 0494

Number of front-end chips338 (barrels) 940 (barrels)

472 (disks) 384 (disks)

Total 810 1324

Number of readout channels asymp 103 middot 106 asymp 17 middot 105

Table 24 Compilation of design parameters for the MVD

MainSub-layer

(rdef) 〈ropt〉 rmin rmax (zdef) 〈zopt〉 zmin zmax

layer [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm]

BarrelInner ring

2522 2180 2275

- - minus398 98layer 1 Outer ring 28 2780 2858

BarrelInner ring

50475 4730 4785


Barrel 95 92 8972 9686 - - minus1338 1390layer 3


DiskSdk 1 front

lt 50 - 1170 3656 20 22197 199

layer 1 Sdk 1 rear 241 243

DiskSdk 1 front

lt 50 - 1170 3656 40 42397 399


DiskLdk 1 front

lt 95 - 1170 7396 70 72697 699

layer 3 Ldk 1 rear 741 743

DiskLdk 2 front

lt 95 - 1170 7396 100 102997 999


Disk

Ldk 3 front- - 1170 7396

160

1501477 1479

layer 5

Ldk 3 rear 1521 1523

StripDk 1 L- - 7433 13115 1625

1600 1603

StripDk 1 S 1650 1653

Disk

Ldk 4 front- - 1170 7396

230

2202177 2179

layer 6

Ldk 4 rear 2221 2223

StripDk 2 S- - 7433 13115 2075

2047 2050

StripDk 2 L 2097 2100

Sdk small pixel disk Ldk large pixel disk StripDk strip disk

Table 25 Positions of the active sensor volumes within the different detector layers rdef and zdef are thepredefined values for the radius and the axial extension as given in figure 24 〈ropt〉 and 〈zopt〉 are the correspondingmean values in the sub-layers after optimisation rminmax and zminmax refer to the outer limits of the radial and axialextension in the respective layers


Plotted results for a silicon thickness of 06 mmand 12 mm represent the minimum effective pathlengths of particles crossing two and four pixel lay-ers respectively With a minimum flight path ofroughly 10 cm between the pixel and the barrel lay-ers the lateral track resolution due to scatteringeffects is restricted to a few hundred micrometersfor particles with a momentum well below 1 GeVc

In total there are three different strip sensor typeswhich have different shapes They are shown infigure 27 on the top Quadratic and rectangularsensors are chosen for the barrel part while a trape-zoidal design is used for the forward disks The re-spective stereo angle is set to 90 in case of barreltype sensors and to a reduced angle of 15 for thetrapezoids In this way strips always run in parallelto one sensor edge The strip pitch for the rect-angular barrel sensors is set to 130 microm A smallervalue of 675 microm for the trapezoids in the disk partis used to recover the worsened resolution inducedby the smaller stereo angle

The strip barrel layers are formed by a paddle-wheelsensor arrangement Neighbouring trapezoidal sen-sors of the forward disks are located at two positionsslightly shifted along the beam axis in order to makeup for the passive sensor edge To accommodate thetarget pipe crossing some barrel strip sensors areleft out at the top and bottom A schematic pictureof the strip sensor geometry is given in the bottompart of figure 27

Finally the defined sensor geometry results in ap-proximately 103 million pixel and roughly 200000strip readout channels Design parameters for theMVD are summarised in table 24 A compilationof the sensor positions in all detector layers can befound in table 25

Figure 29 Schematic routing scheme for the MVDBlue and green arrows illustrate the concept for the bar-rel and the forward part respectively A probable rout-ing of the additional forward disks indicated with yellowarrows potentially interferes with the MVD volume

Figure 210 Detector integration of the MVD insidethe target spectrometer Left Three-point fixation tothe central frame which also provides fixation points forthe beam pipe (a) the target pipe (b) and the connec-tion to a rail system (c) The main MVD parts are inte-grated in two half-frames (right) which are connectedseparately to the central frame

Mechanics and Detector Integration

The MVD will operate under normal pressureconditions at room temperature of approximately+25 C An active cooling of both pixel and stripreadout electronics is needed for a long term sta-bility of the detector A detailed description of thecooling system can be found in chapter 56 Besidesthe cooling pipes other service structures must bebrought from the outside close to the silicon detec-tors in order to supply sensors and associated elec-tronics components as well as to transmit signalsfor data and slow control of the readout chips Aschematic routing scheme for all MVD services isgiven in figure 29

Limitations are given by the stringent boundaryconditions Due to the fixed target setup whichdefines the physically most interesting region in theforward part a guiding to the outside of the target


spectrometer is only possible in upstream directionIn this region the opening cone of the beam pipe re-duces the available cross section at inner radii Therestricted radial space requires a circular occupancyof all services around the beam pipe For the pixeldisk inside the barrel layers a bundling at the topand bottom is foreseen Patch panels which allowa connection to peripheral systems are intended tobe placed further upstream

Basic ideas of the overall detector integration forthe MVD are shown in figure 210 The mechanicalinterface to external parts of the PANDA apparatusis defined by a three point fixation to the centralframe which serves as reference frame and globalsupport for the entire central tracking system Italso facilitates a consecutive assembly of the beam-target system the MVD and the outer tracker out-side of the solenoid magnet A safe insertion of allattached sub-systems to their nominal position isrealised via a rail system The global support struc-ture for the MVD is composed of two half-framesBoth halves are mechanically independent in orderto avoid any introduction of stress to either half-detector They allow the internal mounting of alldetector half-layers and the attachment to the cen-tral frame All support structures inside the MVDwill be composed of light-weight carbon materialsDetails can be found in section 55 Additional sup-port structures are also needed for MVD services

References

[1] K Nakamura et al (Particle Data Group) Re-view of particle physics J Phys G 370750212010

[2] A Capella et al Dual parton model PHYSICSREPORTS 236(4amp5)225ndash329 1994

[3] M Bleicher et al Relativistic hadron-hadron collisions in the ultra-relativistic quan-tum molecular dynamics Nucl Part Phys251859ndash1896 1999

[4] F Hugging Development of a Micro-Vertex-Detector for the PANDA-Experiment at theFAIR facility IEEE Nuclear Science SymposiumConference Record N30(190)1239ndash1243 2006

[5] Th Wurschig Design Optimization of thePANDA Micro-Vertex-Detector for High Per-formance Spectroscopy in the Charm QuarkSector PhD thesis Universitat Bonnurnnbndehbz5N-26230 2011

[6] GR Lynch and OI Dahl Approximationsto multiple Coulomb scattering Nucl InstrMeth B586sndash10s 1991


35

3 Silicon Pixel Part

This chapter describes the silicon pixel part of theMicro Vertex Detector The innermost layers anddisks of the MVD are the most important onesfor determining primary and secondary vertices ofcharm mesons They need a very high precisionand granularity to cope with the required spatialresolution In addition they have to handle hightrack densities in the forward region for proton tar-gets they are exposed to relatively high radiationlevels and they have to process and transmit veryhigh data rates due to the untriggered readout ofPANDA To cope with these requirements a hybridpixel detector with its two-dimensional segmenta-tion featuring a 2D readout without ambiguity dou-ble hit resolution excellent signal to noise ratio athigh speed and reliable sensor technology appearsto be the most interesting choice for equipping theinnermost part of the MVD

31 Hybrid Pixel AssemblyConcept

Hybrid pixel detectors have been studied and de-veloped for the vertex trackers of the LHC experi-ment The core issue is a two-dimensional matrixthe sensor (rectangular shape of few tens of mmby a few hundreds microm thick) of reverse biased sili-con diodes flip-chip bonded to several readout chipsEach cell (pixel) on the sensor matrix is connectedby solder-lead bumps (alternatively indium is used)to a frontal cell on a readout ASIC developed inCMOS technology The readout cell must fit intothe same area as the pixel sensing element In par-ticular the ASICs used in the LHC experimentshave been developed in 250 nm CMOS technologywhereas the sensor has been made of Floating Zone(FZ) silicon with a minimum thickness of 200 micromThe request of high granularity and the simulationresults of benchmark channels for the MVD indi-cate a pixel size of 100 micromtimes 100 microm as the mostsuitable However the reduction of the surface ofthe pixel increases the perimeter-to-area ratio andhence the input capacitance which is usually dom-inated by the inter-cell capacitance To cope withthe limited material budget an RampD effort to inves-tigate a lower sensor thickness has been made andthe results are reported in chapter 32 Concern-ing the sensor material references [1] [2] indicatethat silicon epitaxial material could be an interest-

ing material to be applied in PANDA in particularfor radiation hardness up to the radiation level ex-pected Taking into account the configuration ofthe wafer made of an epitaxial silicon layer grownon top of a Czochralski substrate thinned pixel sen-sors could be obtained by removing most of theCz substrate Furthermore an RampD project hasstarted to develop a readout ASIC in 130 nm CMOStechnology equipped with pixel cells of the requiredsize In addition to the two-dimensional spatial in-formation time and energy loss of the hit will bemeasured Besides the feature to obtain by thistechnology the same electrical functionality using aquarter of area than older technology with a powerreduced by a factor 4 appears a challenge towardsa simplification of the cooling system The readoutarchitecture is described in the section 333 and theASIC prototypes in the section 335 A schematicview of the custom hybrid pixel assembly for thePANDA experiment is shown in figure 31

Figure 31 Scheme of the hybrid pixel for PANDA

The epitaxial silicon layer with a thickness up to150 microm and a high resistivity of few kOhm cm isthe active region of the pixel matrix The Cz sub-strate features lower resistivity (001minus 002 Ω middot cm)and a thickness of a few hundreds of microm Mostof this substrate is removed starting from the sideopposite to the epitaxial layer Each pixel is con-nected by the bump bonding technique to the cor-responding readout cell of a custom ASIC that willbe developed in 130 nm CMOS technology Thenthis assembly is glued to a carbon foam layer toimprove heat dissipation towards the cooling pipeas a part of a cooling system working in depressionmode and based on water as cooling fluid A struc-


ture made of carbon fibre with suitable shape is themechanical support

32 Sensor

321 First Thinned Prototypes

To study the epitaxial silicon material first proto-types of hybrid pixel assemblies using thinned epi-taxial silicon pixel structures as sensor element havebeen produced and tested in 2008 [3] Results wereobtained from assemblies with three different epi-taxial thicknesses of 50 75 and 100 microm The front-end chip for PANDA was still under developmentat that time and therefore ALICE pixel front-endchips and readout chain were used Basic detectorcharacteristics bump bond yields were measuredand tests with various radioactive sources were per-formed

Design and Production

For the sensors 4-inch wafers with three differentn-epitaxial layer thicknesses deposited on standardCzochralski (Cz) substrate wafers were provided byITME (Warsaw) The different epitaxial layer char-acteristics are shown in table 31 The Cz sub-strate parameters were equal for all wafers n+

conductivity type Sb dopant 525 microm thickness001minus 002 Ω middot cm resistivity

Epi Layer Thickness Resistivity

Epi-50 49plusmn 05 microm 4060 Ω middot cm



Table 31 Thickness and resistivity of the epitaxiallayers of the first assemblies

The wafers were processed by FBK-ITC (Trento)using ALICE pixel sensor masks provided by INFN-Ferrara which had already previously been used toproduce high quality ALICE pixel assemblies from200 microm Float Zone (FZ) material [4] The sensormasks include full-size and single-chip sensors (theALICE pixel size is 50 micromtimes 425 microm) together withthe usual range of diagnostic structures includingsimple diodes Six epitaxial wafers were processedtwo of each thickness The thinning of the sensorwafers and the bump-bonding to ALICE pixel read-out chips were performed by VTT (Finland) Thetarget thicknesses for the sensor wafers were chosento be 100 120 and 150 microm respectively for the50 75 and 100 microm epi layers Single chip sensors

were bump-bonded to chip-sensor assemblies (Epi-50 Epi-75 and Epi-100 respectively correspondingto the 50 75 and 100 microm thick epitaxial layers)

Results

The full-depletion voltage values for all three epi-taxial wafer thicknesses were less than 6 V Fig-ure 32 shows the total sensor leakage current as afunction of bias voltage up to 200 V for two assem-blies

Figure 32 Leakage current as a function of the biasvoltage for assemblies Epi-50 (lower curve) and Epi-75(upper curve)

No sign of breakdown is observed and the thickersensor has a higher current for any given voltageGood long-term stability was also noted Totalleakage current as a function of the bias voltage wasmeasured at different temperatures up to 50 C forone assembly (Epi-50) A leakage current of about3 nA corresponding to a bias voltage of 100 V (theassembly has 8192 pixels) at 21 C increases up toabout 23 nA at 43 C This corresponds to a leakagecurrent per pixel of less than about 10 fA which isa value that can be accepted and controlled by thereadout circuit

Three assemblies S15 (Epi-50) S8 (Epi-75) and S6(Epi-100) were instead mounted on testing boardsenabling their complete electrical characterisationusing the ALICE Pixel DAQ system [5] Beta par-ticles from a 90Sr source which pass through theassemblies and their associated mountings to subse-quently trigger a plastic scintillator with photomul-tiplier readout should be sufficiently similar in theirbehaviour to minimum-ionising particles (MIP) toenable a relative comparison of the active thicknessof the epitaxial sensors Calibration of the thresh-old DAC setting in electrons is obtained by mak-ing similar but selfndashtriggered threshold scans whileexposing the assemblies to 55Fe and 109Cd Xndashraysources Figure 33 shows the threshold value inelectrons corresponding to the most probable value


of the Landau distribution plotted with respect tothe nominal epitaxial active thickness

Figure 33 Threshold values corresponding to theLandau most probable value for the different epitaxiallayer thicknesses

It should be noted as is confirmed by the linearityof the result that charge sharing is not a major is-sue despite the 50 microm pixel width due to the factthat the active thickness is equally limited Taking22500 electrons for a true MIP in 300 microm of siliconone would expect 7500 electrons from 100 microm ofsilicon and given the use of the beta source whichshould raise that value somewhat the agreementseems more than satisfactory

322 Radiation Damage Study

Epitaxial material has been studied for radiationdamage by many experimental groups (RD50 col-laboration) In particular at the beginning of theRampD phase for the PANDA pixel detector resultson thin epitaxial diodeds (about 25 microm) and lowerepitaxial resistivity (about 100 minus 150 Ω middot cm) wereavailable However the application of epitaxial ma-terial in the PANDA experiment asked for higherthickness and higher resistivity to obtain more elec-tron production from MIP and a lower full depletionvoltage respectively For these reasons systematictests for radiation damage were planned The mostimportant parameters as full depletion voltage andleakage current have been studied as a function offluences and annealing phases To study the be-haviour of these parameters displacement damagetests were performed using neutrons from the nu-clear reactor (TRIGA MARK II) of Pavia (Italy)at the LENA laboratory Diodes from wafers fea-turing different epitaxial thicknesses and resisitivityas reported in Table 32 have been characterisedfor full depletion voltage and leakage current us-

ing a probe station with bias voltage provided by aKeithley 237 High Voltage Source Measure Unit

Diodes Thickness Resistivity

Epi-50 HR 49plusmn 05 microm 4060 Ω middot cm



Epi-50 MR 50plusmn 1 microm 3100 Ω middot cm



Epi-75 LR 745plusmn 1 microm 460 Ω middot cm

Table 32 Thickness and resistivity featuring the epi-taxial layer of the wafers used for the radiation damagetests HR MR and LR mean High resistivity MiddleResistivity and Low Resistivity respectively

I-V measurements have been performed using anHP4145B Semiconductor Parameter Analyser andC-V mesurements have been performed using anHP4248A Precision Meter The measurents of I-V and C-V trends for different diodes have beencarried out before and after the irradiation as afunction of the equivalent fluence and specificallywith the following values 15 middot 1014 and 53 middot1014 n1MeV eq cmminus2 respectively corresponding toabout 10 years of PANDA lifetime with antiprotonproton annihilation and antiproton Xe interactions(50 duty cycle)

Then a long annealing phase at 60 C was startedand full depletion bias voltage and leakage currentcharacterisation were performed many times dur-ing this period The full depletion voltage is evalu-ated from the intersection of the two extrapolatedlines in the capacitance-voltage characteristic InTable 33 the full depletion voltage of diodes undertest are reported before the irradiation

Diodes Full Dep Voltage [V] σ [V]

Epi-50 HR 435 006

Epi-75 HR 56 01

Epi-100 HR 59 02

Epi-50 MR 491 007

Epi-75 MR 82 03

Epi-100 MR 104 06

Epi-75 LR 426 13

Table 33 Pre irradiation full depletion voltage valuesas measured for epitaxial diodes under test for radiationdamage HR MR and LR mean High resistivity MiddleResistivity and Low Resistivity respectively σ is thestandard deviation

As expected the full depletion voltage values de-crease as the resistivity of the epitaxial part in-creases In figure 34 the full depletion bias


Figure 34 Annealing behaviour of the full depletion voltages as a function of the annealing time for the diodesirradiated to 15 middot 1014 n1MeV eq cmminus2 All bias voltage values are normalised to the 50 microm thickness The valuesreported at time zero are the voltages of irradiated diodes measured after the irradiation but before the annealingphase In this plot the square corresponds to diodes Epi-50 the circle to diodes Epi-75 and the triangle to diodesEpi-100 the colours in red scale correspond to the high resistivity devices the colours in blue scale to the middleresistivity diodes and the green to the low resistivity one

voltages measured during the annealing phase at60 C following the irradiation corresponding to15 middot 1014 n1MeV eq cmminus2 All voltage values are nor-malised to the 50 microm epitaxial thickness The biasvoltages reported at time zero are the values mea-sured after the neutron irradiation and before theannealing phase [6]

The measured values of the full depletion voltagesfor high and middle resistivity diodes have increas-ing trends showing a reverse annealing effect Prac-tically the devices have undergone the type inver-sion phenomenon immediately after the neutron ir-radiation Besides the larger epitaxial thicknessneeds higher full depletion voltages that could beexplained both by higher resistivity values of theepitaxial material or by a non-uniform concentra-tion of oxygen diffusing from the Czochralski sub-strate limiting its beneficial action against the ef-fects induced by radiation The full depletion volt-age values for the lowest resistivity diodes as a func-tion of the annealing time show a different trend incomparison to the others The evidence for typeinversion is found after few days of the annealingphase

The diode volume currents corresponding to the fulldepletion bias voltage of the devices under study in-crease to values of about 10minus2 Acm3 for all testeddevices starting from 10minus8 minus 10minus7 Acm3 A leak-age current of about 13 nApixel corresponding tothe maximum measured volume current assuming avolume of 100 micromtimes 100 micromtimes 100 microm the PANDAchoice is reached Since the front-end electron-ics has a leakage compensation scheme designed towithstand a leakage current of 60 nA with a base-line shift smaller than 2 mV this increase of leakagecurrent is already acceptable for the PANDA exper-iment The leakage current decreases to 50 aftersome days of annealing The damage constant αranges between 5 minus 8 middot 10minus17 Acm This constantis estimated as ∆Jφ where ∆J is the differencebetween the diode volume current before and afterthe irradiation both measured once the full deple-tion bias voltage has been reached and φ is thecorresponding equivalent fluence

Results from a test at 53 middot 1014 n1MeV eq cmminus2 inparticular for the annealing phase at 60 C confirmthe trends already discussed but with higher fulldepletion voltage values corresponding to the same


time of annealing The diode volume currents afterthe irradiation are of the same order of magnitude

Comparison of full depletion voltage values as afunction of two different temperatures in the an-nealing phase (40 C and 60 C) shows differenttrends evidence of inversion type is not limitedto the lowest resistivity under test but even in thehigher resistivity samples Apart from that the fulldepletion voltages are systematically lower at lowertemperature as expected

Looking at these results the lower thickness appearsbetter for an application as sensor but this param-eter has to be a compromise with the dEdx in-formation and the signal to noise ratio Besides ahigher thickness entails larger stiffness The 100 micromthickness appears to be the best compromiseThe resistivity choice has to be a compromise be-tween the initial full depletion voltage values andthe values measured during the annealing phaseThe resistivity for the epitaxial layer can be ad-justed in a certain range between 1 kΩ middot cm and3 kΩ middot cm

323 Fullsize Prototype Sensors

A first design of the full size sensors for the pixeldetector has been carried out taking into accountthat sensors of different sizes are needed for opti-mising the disk coverage


The basic idea for the PANDA pixel detector is amodular concept based on the readout chip sizeSensors housing 2 4 5 or 6 readout chips have beenarranged in the epitaxial silicon wafer to allow afirst production for thinning and dicing studies Infigure 35 the picture of a whole wafer with pixelsensors and diagnostic structures on the edge canbe seen

Figure 36 shows a partial view of the wafer In themiddle two sensors for arranging six readout chipsare visible The vertical white lines correspond totwo larger pixel (100 micromtimes 300 microm) columns in theregion where two readout ASICs will be arrangedside by side

Figure 37 represents a partial view of a sensor madewith 100 micromtimes 100 microm pixels At one corner of eachpixel the pad for the bump bonding is visible Theseare arranged in a mirror configuration with respectto a bus serving two pixel columns The pitch alongthe column is 100 microm and along the rows is alter-natively 50 microm and 150 microm

Figure 35 Picture of a PANDA wafer

Figure 36 Picture showing a partial view of the sen-sor wafer

Figure 37 Photograph of a part of the pixel matrixPads for bumps are the white circles at one corner ofthe gray pixel


In these sensors the passive silicon edge housingguard rings is 700 microm wide (about 10 of the sen-sor size) The guard ring closest to the sensor hasa width of 100 microm and it is connected to groundadditional guard rings are floating This configura-tion has been applied as a preliminary layout sim-ilar to the previous sensors designed for LHC ex-periments In our case the use of epitaxial siliconsensors gives the possibility of a reduced bias volt-age during the sensor lifetime as a consequence ofa lower full depletion voltage A limited guard ringnumber reducing the previous passive silicon edgeof 50 could be possible This evaluation has beenobtained from simulations carried out in collabora-tion with FBK showing the break down voltage as afunction of the floating guard ring number Experi-mental measurement using the single chip assemblybased on ToPix v3 will be performed to validate thesimulation resultsMiddle resistivity wafers used during the radiationdamage test have been investigated for these testsThree epitaxial thicknesses were chosen to better in-vestigate the thickness dependence during the thin-ning process Epi-75 low resistivity wafers wereadded for comparison purposes Sixteen waferswere thinned at the Wafer Solution company theEpi-50 and Epi-75 wafers to the target of 100 micromthicknesses and the Epi-100 wafer to 120 microm thick-ness in addition half of these wafers were diced toobtain sensors No mechanical breakage occurred

Tests

In addition to the radiation damage test of the epi-taxial material used to make the first sensor pro-totypes with the results already described in theprevious section I-V and C-V measurements andplanarity and thickness tests were performed

In particular I-V and C-V measurements were madeon some test structures before and after thinningand dicing of wafers to investigate potential me-chanical damage in the bulk of the wafer Variationswithin measurement errors have been reported inthe I-V and C-V trends and no changes in the leak-age current and full depletion voltage values wereobserved

Concerning the planarity evaluation two tech-niques were used compatible with the equipmentat disposal at INFN-Torino a Mitutoyo measure-ment machine equipped with both a precision CCDcamera (magnifier 110) and a stylus probe (3 micromover 1 m precision) Simply to put down the 4 inchthinned wafers on the measurement base and usinga CCD camera a total unplanarity of about 40

was measured by a grid of 8 times 9 measurement po-sitions and a preferential cup shape was observedThis result can be explained both with the elastic-ity of the wafer now thinned and a measurementimprecision due to the reflection of the light on themirror like wafer surface

By using the probe the elasticity of the wafer wasimmediately detected by observing the movementof the wafer under the probe pressure without anybreakage The thickness measurement has been ob-tained using the same grid of 72 measurement posi-tions as previously discussed Figure 38 shows theresult obtained for a 100 microm thick nominal wafer

Figure 38 Plot showing the unplanarity of a thinnedwafer prototype measured with a stylus probe using a72 position grid

324 Technology Choice Epi vsOxygen

The higher radiation tolerance of the epitaxial de-vices in contrast to FZ ones can be explained takinginto account the oxygen diffusion from the Cz sub-strate [7] as shown in figure 39

Figure 39 Oxygen concentration as a function ofwafer depth for different epitaxial thicknesses [7]


According to the oxidation process used to increaseoxygen concentration in the FZ wafers [7] as shownin figure 310 some epitaxial wafers were submittedto an oxygen process at the FBK In particular a1100 C phase for 12 hours with wafers in oxygenatmosphere was followed by a phase of 536 hoursat 1100 C temperature in a nitrogen atmosphereThen SIMS measurements of the oxygen concentra-tion as a funtion of depth were performed at ITME

Figure 310 Oxygen concentration as a function ofFZ wafer depth for some oxidation processes [7]

Figure 311 reports the result for an epitaxial wafer100 microm thick The flat trend can be explained withboth contributions coming from the Cz substrateoxygen diffusion and from diffusion from oxygen at-mosphere during the high temperature phase Sucha result shows the possibility to have a uniform tol-erance to the radiation along the sensor bulk

Figure 311 Oxygen concentration as a function ofepitaxial wafer depth The scatter graph represents thevalue of oxygen concentration measured on beveled sur-face The magenta line shows the concentration of oxy-gen in shallow depths at sample ion etching

325 Production

Taking into account the first design of the full sen-sors for the pixel detector as reported in the draw-ing of figure 312 and the needed sensor number formaking the whole pixel detector as reported in table34 a minimum number of 34 wafers has to be pro-duced More details will be explained in the section36

Figure 312 Scheme of the wafer with the first designof the full size sensors

S2 S4 S5 S6 Total

Barrel1 6 8 0 0 14

Barrel2 0 0 6 44 50

Disk1 6 2 0 0 8

Disk2 6 2 0 0 8

Disk3 4 4 12 4 24

Disk4 4 4 12 4 24

Disk5 4 4 12 4 24

Disk6 4 4 12 4 24

Total 34 28 54 60 176

Chip Number 68 112 270 360 810

Table 34 Sensor numbers S2 is the sensor with tworeadout chips S4 with four S5 with 5 and S6 with 6sensors respectively

In particular quality control of each patternedwafer will be performed on test structures adjacentto the sensors C-V I-V trends on test diodes C-V on MOS structure I-V trend on a gated diodebreakdown voltage for several GR (Guard Ring)configurations C-V interpixel measurement Visualinspection of metals and bonding pads of each sen-sor can be planned for checking for example theetching uniformity Doping profiles as a function ofepitaxial depth and bias voltage are additional mea-surements useful to compare a wafer before and af-ter the oxygen process Besides wafer samples fromproduction processes will be used to obtain diodes


for radiation damage tests to monitor the radiationresistance uniformity

ITME has been selected as the company able toprovide the epitaxial wafers useful for our project bychecking samples during the RampD FBK has shownthe capabilities for sensor production

In table 35 the most probable properties of the epi-taxial wafer for the production are summarised

Cz substrate

Diameter 100plusmn 05 mm

Orientation lt 100 gt

Conductivity type n+

Dopant Sb

Thickness 525plusmn 20 microm

Resistivity 001minus 002 Ω middot cm

Epitaxial Layer

Conductivity type n

Dopant P

Thickness(centre) 100plusmn 2 microm


Table 35 Epitaxial wafer properties for sensor pro-duction

33 Front-End Electronics

331 Requirements

The pixel readout ASIC has to provide for each hita simultaneous position time and energy loss mea-surement The specifications for the readout elec-tronics depend on the requirements of the MVDThe fundamental parameters to take into accountare the PANDA radiation environment and the closeproximity of the MVD to the interaction pointA pixel size of 100 micromtimes 100 microm has been chosenbased on the study of track performance Sinceeach readout cell has to be bump bonded to thesensor the readout cell matrix has to respect thesame geometry as the pixel sensor matrix There-fore the readout cell has to fit in an area of100 micromtimes 100 microm To guarantee the reliability ofthe system time and charge digitisation has to beperformed at the pixel cell level in order to transmitonly digital information out of the pixel cell For aMinimum Ionising Particle (MIP) in a 100 microm thickepitaxial sensor an average ionisation charge of sim13 fC is expected The upper limit of the dynamicrange is given by the measurement of protons witha momentum down to sim 200 MeVc which can de-posit ionisation charges up to 50 fC A power den-

sity of up to Wcm2 can be tolerated However astrong effort is going into the reduction of the powerconsumption of the front-end electronics and hencethe material budget for the cooling systemIt is foreseen to employ a p-in-n epitaxial sensorwhich should give an adequate radiation hardnesswith a simple geometry However the possibilityof using the already well known n-in-n sensor em-ployed in LHC is considered as a backup solutionTherefore the front-end has to be designed to becompatible with sensors of either polarity Thenoise level is mainly limited by the input transis-tor thermal noise and therefore is a function of itsbias current An Equivalent Noise Charge (ENC) ofsim 200 electrons is a good trade-off between powerconsumption and minimum detectable charge Thereadout electronics has to work in the PANDA radi-ation environment with an accumulated Total Ion-ising Dose (TID) of 100 kGy in 10 years Hit ratesimulation studies estimate an average rate of max-imal sim 103 events per second on a single pixel cellThe system will be self-triggered and all the eventsabove a given threshold have to be read out lead-ing to a high data rate A time resolution lowerthan the average event rate of 2 times 107 eventss isrequired Table 36 summarises the specificationsfor the ASIC design

Pixel size 100 micromtimes 100 microm

Self trigger capability

Chip active area 114 mmtimes 116 mm

116 rows 110 columns

Chip size 116 mmtimes 148 mm

dEdx measurement ToT

Energy loss 12 bits resolution

measurement 7 bits linearity

Input range up to 50 fC

Noise floor le 0032 fC

Clock frequency 15552 MHz

Time resolution 645 ns (19 ns rms)

Power budget lt 800 mWcm2

Max event ratecm2 61times 106 Hitss

Max data ratechip sim 250 Mbits

Total ionising dose le 100 kGy

Table 36 Specification summary for the front-endASIC ToT Time over Threshold

332 Readout Architecture

The readout architecture depicted in figure 313foresees a module structure formed by a three-layerstructure the sensor the front-end chip and the


Figure 313 Pixel readout scheme

transmission bus The sensor is bump-bonded tothe readout chips which in turn are wire-bondedto the transmission busThe module is connected to a service board locatedoutside the sensitive area which performs the datamultiplexing and the electrical-to-optical conversionto decrease the amount of cabling The link is bi-directional and allows to send the clock and to up-load configuration parameters to the front-end Asecond service board hosts the voltage regulatorsDue to the low power supply voltage of the front-end (12 V) a DC-DC down-conversion will be im-plemented at this level

As shown in figure 313 two possible solutionsare under evaluation in the first one a modulecontroller chip is used on top of the transmissionbus to receive multiplex and re-transmit the datafrom the front-end chips over high speed seriallinks In the second solution the module is directlyconnected to the service board The advantage ofthe first solution is the reduction of the numberof wires coming out the module moreover bymultiplexing the data from various front-end chipsa better bandwidth allocation is possible On theother hand when compared to the second solutionit requires an extra ASIC which increases the powerconsumption and becomes a critical point of failure

333 Front-End ASIC

A new front-end ASIC named ToPix is underdevelopment in order to cope with the uniquerequirements of the PANDA experiment The cir-cuit is developed in a commercial CMOS 013 microm

technology in order to profit from the high levelof integration and the intrinsic radiation toleranceof deep submicron technologies Indeed it hasbeen shown [8] that deep submicron technologieswith very thin gate oxide and Shallow TrenchInsulation (STI) have the potential to stand highlevels of total ionising dose with standard layouttechniques In order to cope with the Single EventUpset (SEU) effect all flip flops and latches in thepixel cell are fully static and are based either onHamming encoding or triple redundancy

Pixel Cell

In the proposed design the Time over Threshold(ToT) technique has been adopted for the ampli-tude measurement The ToT allows to achievegood linearity and excellent resolution even whenthe preamplifier is saturated thus making room fora high dynamic range Owing to the saturation ofthe input stage a cross talk of the order of few per-cent is expected for large input charges The ToTis obtained by taking advantage of the time stampwhich is already distributed to all pixels in order toobtain the time information

Figures 314 and 315 show the schematic of theanalogue and digital parts of the pixel cell respec-tively The charge preamplifier is a gain-enhanceddirect cascode amplifier with a feedback capacitorand a constant current discharge circuit A lowfrequency differential amplifier in feedback with theinput stage allows for leakage current compensationby forcing the stage DC output value to a referencevoltage The input amplifier can be configured to


Figure 314 Analog readout channel

Figure 315 Pixel control logic

accept signals of both polarities via control bitsA calibration signal can be applied to the input ofeach preamplifier via a 24 fF integrated capacitorEach pixel can be independently enabled to receivethe calibration signal via the configuration registerThe comparator is based on a folded cascodearchitecture The threshold voltage is set by globalDAC and can be fine adjusted via a 4 bits localDACWhen the preamplifier output crosses a thresholdthe comparator switches to 1 and the time stamp

value is loaded into the leading edge register Theintegrating capacitor is then slowly dischargedand when the amplifier output goes below thethreshold the comparator switches back to 0 andthe time stamp value is loaded into the trailing edgeregister Therefore the time stamp value in theleading edge register gives the timing informationwhile the difference between leading and trailingedge time stamps gives the amplitude information


Figure 316 Column readout schematic

ToPix Architecture

The pixel cells are organised in double columnswhich share the same time stamp and readoutbuses Each double column is composed of 2times 116cells a 12 bit time stamp bus a 7 bit address busand a 12 bit data bus Each cell in the columnhas a unique cell address which is hardwired Thereadout buses are differential The time stampbus uses a reduced swing pseudo differential logicwith pre-emphasis to limit the power consumptionwithout degrading the speed performance Thetime information is sent on the bus with a Grayencoding to reduce the switching noise and to avoidsynchronisation problems with the asynchronouspixel logic The address and data readout buses areCMOS pseudo differential lines and are read out atthe end of the column via sense amplifier in orderto keep the cell output drivers to a manageablesize and power consumptionThe column readout scheme is based on a conceptsimilar to the one used in other pixel readoutdesigns [9] When a pixel cell receives a hit itgenerates a busy signal which is put in logical OR

with the busy of the previous cell and then sent tothe next cell Therefore at the end of the columnthe busy signal is high if at least one pixel containsdataThe readout is made in a fixed priority scheme theread commands are sent to all pixels but only theone with its busy input signal at zero (ie with nopixel at higher priority with data) sends its dataand address to the output bus A freeze signal isused to block the generation of new busy signals(but not the possibility to store an event into thepixel) during readout in order to make sure thatall the pixels are addressed in a readout cycle

The configuration data upload uses the samescheme the config mode signal puts all pixels inthe busy state Successive commands address thepixels from the one with highest priority to theone with the lowest During each address phaseit is possible to write the data on the time stampbus (now used to send configuration informationand thus not connected to the counter) in theconfiguration register or to put the data stored inthe configuration register on the data bus


Figure 317 Frame structure CRC Cyclic Redundant Check ECC Error Correction Code

Each double column is controlled by a ColumnReadout Control Unit (CRCU) The CRCU con-sists of four main components a bank of differentialdrivers for the time stamp propagation a bank ofsense amplifiers for the data readout a 32 cells out-put FIFO and the control logic The CRCU sendsthe readout signals on the base of the state of thebusy from the pixel column and the state of theoutput FIFOThe ToPix readout is controlled by a Chip Con-troller Unit (CCU) which reads the data from the55 CRCUs and multiplexes them to up to 3 doubledata rate serial links The multiplexing operationis made in a round robin scheme with two queues afast one for the CRCU FIFOs with less than 5 freecells and a slow one for the othersData are transmitted in frames A frame containsall the data received by the chip during a singletime stamp counter cycle Each frame is dividedfrom the previous and the next ones by a headerand a trailer respectively Figure 317 shows theframe structure

Frames are composed by 40 bits wide words with a2 bit header to identify a data word from a headeror a trailer word The data word contains 14 bitsfor the pixel address and 24 bits for the leading andtrailing edge time information The header wordcontains the chip address the frame counter anderror detection and correction bits The trailer wordcontains the number of words present in the framethe Cyclic Redundant Check (CRC) of the entireframe and the error detection and correction bits

334 Module Controller

One of the two readout options for the pixel sensorsshown in figure 313 foresees a module controllerASIC located on the module structure The useof a module controller has been a common choicein the electronic readout of the LHC pixel detectors[10][11][12] because it gives a greater degree of flex-ibility

In the PANDA MVD readout scheme the use of amodule controller would allow a reduction of thetransmission lines by using a Phase Locked Loop(PLL) and therefore the possibility to transmit dataat a bit rate faster than the clock frequency More-over it would also allow a better load balancingbetween front-end chips with high and low eventratesHowever the disadvantages in terms of higherpower consumption increased number of coolingpipes and increased system fragility due to the in-troduction of a critical single point of failure makesthe solution without module controller more attrac-tive

335 ASIC Prototypes

An RampD activity for the design of the ToPix ASICis ongoing In the following the three ToPix proto-types will be described together with the test re-sults

ToPix Version 1

A small prototype with the basic pixel cell buildingblocks was designed in order to evaluate the ana-logue performances and gain experience with thetechnologyThis first prototype [13] chip contains 32 read-out cells each one equipped with a preamplifierand a discriminator The preamplifier consist ofa Charge Sensitive Amplifier (CSA) with a feed-back capacitance of 10 fF a constant current feed-back and a baseline restorer The amplifier sizeis 37 micromtimes 51 microm and the discriminator size is128 micromtimes 48 microm Figure 318 shows the ToPix v1architecture The ASIC size is 2 mmtimes 1 mm Thecells are arranged in 8 groups of 4 cells Each grouphas 4 calibration input lines 4 analogue outputlines and 4 digital output lines Each input cal-ibration line is in common for one of the cells ineach of the 8 groups In each cell a calibration ca-


Figure 318 ToPix v1 simplified schematic

pacitor of 30 fF shunts the calibration line to theinput node Therefore sending a voltage step onthe calibration line a delta shaped current pulse isinjected at the input node The output lines aremultiplexed in order to be shared between the cellsbelonging to different groups In order to drive thecapacitive load of output pads and measurementprobes which can be up to 10 pF the analogue sig-nals are driven by a source follower and the digitalsignals are driven by a buffer Six cells belonging totwo different groups have direct connection to ded-icated input pads This configuration allows theobservation of the cross-channel effects

Figure 319 shows a simplified schematic view ofToPix v1 readout which is designed to work with ann-type sensor (negative current pulses) The CSAcore amplifier has three feedback components

bull The feedback capacitance CFB fixes the chargegain (1CFB) On the one hand a high valueof CFB decreases the CSA charge gain on theother hand a low value of CFB decreases theCSA loop stability In this prototype a nomi-nal value of the feedback capacitance of 10 fFhas been chosen as the optimal compromise be-tween gain and loop stability

bull The baseline restorer is implemented with agmC low pass filter stage which computes thedifference between the baseline reference volt-age and the CSA output The resulting lowfrequency voltage signal controls the gate of aPMOS which generates the leakage compensa-tion current at the input node

bull The discharging current generator is based ona differential pair The two inputs are con-nected to the CSA output and to the baseline

reference voltage respectively The differen-tial pair is biased by a current source WhenVOUT = VREF the current is equally shared inthe two branches of the differential pair Whena negative current signal is present at the CSAinput VOUT increases thus switching off oneof the two branches Consequently the cur-rent flows in the input node and CFB is dis-charged A good constant discharge is achievedonly when VOUT is sufficiently large to fullysteer the current from one branch to the otherof the differential pair Since the two transis-tors at the differential pair input work in weakinversion this voltage is about 40 mV

Figure 319 ToPix v1 cell schematic

The constant current discharge provides a constantslope trailing edge of the preamplifier output sig-nal which in turn provides a linear ToT measure-ment of the sensor charge signalThe discharge cur-rent source (IFB) is normally turned off When an


input signal is applied the output voltage of thepreamplifier increases very fast exceeding the volt-age VREF P the current source is turned on andthe feedback capacitor is discharged with constantcurrent until equilibrium is restored and the outputlevel of the amplifier turns back to VREF P Thedynamic range is extended by discharging the feed-back capacitor at the input node preserving thelinearity of the ToT measurement even for higherinput signals when the preamplifier output is sat-urated The feedback topology considered is simi-lar to the one employed in [9] The discriminatordesign follows a low-voltage approach employing asingle stage folded-cascode topology followed by twoinverters For test purposes each channel is pro-vided with an internal calibration capacitor CCAL

driven by an external signal INCAL with amplitudeVCAL The injected charge is given by the productVCAL middotCCAL The absolute value of CCAL is knownwith an accuracy of 10

ToPix Version 1 Test Results

Figure 320 shows a typical pulse at the outputof the preamplifier for an input charge signal of05 fC measured with an oscilloscope and a dif-ferential probe for best common mode noise rejec-tion The preamplifier gain in the linear region is60 mVfC This value was obtained fitting 5 cal-ibration points with nominal values of 1 3 4 812 fC The rms output noise is 1 mV Assuminga gain of 60 mVfC this translates to an equivalentinput noise of 100 eminus rms The power consumptionis 12 microW per pixel cell using a 12 V power supplyThe test results are summarised in table 37

Figure 320 Cell output for an input charge of 05 fC

ToPix Version 2

A second prototype of the ToPix ASIC has beendesigned and tested in order to prove the pixel cellsand column functionality This chip is designed for

analogue gain 60 mVfC

Linear input gain 100 fC

Vnoiserms (Cd = 0) 1 mV

ENC (Cd = 0) 104 eminus

Power consumption 12 microWcell

ToT dispersion (rms) 20

Baseline dispersion (rms) 13 mV

Table 37 ToPix v1 electrical tests result

a clock frequency of 50 MHz as in the original spec-ifications

The prototype includes two 128-cell and two 32-cell columns The pixel cells include all the func-tional blocks required for the final ASIC except forthe bonding pad opening The end-of-column logichas been greatly simplified for test purposes and in-cludes only the counter and the bus readout senseamplifiers The prototype block diagram is shownin figure 321


The chip has been designed in a commercial CMOS013 microm technology with 8 metal layers The two128-cell columns are folded in four 32-cell slices inorder to have an acceptable form factor This ar-rangement represents a worst case with respect tothe final floorplan because of the longer intercon-nections The die size is 5 mmtimes 2 mm Figure 322shows the chip die

The Single Event Upset (SEU) effect has beenmitigated by designing all flip flops and latchesin the pixel cell as fully static and based on theDICE architecture [14] This architecture shownin figure 323 is insensitive to SEU as far as onlyone circuit node received the voltage spike due to


Figure 322 ToPix v2 die

the incoming particle The area penalty is about afactor of 2 compared with the factor 35 minus 4 of atriple redundancy circuit

Figure 323 Dice cell schematic

A number of features has been added in order to im-prove testability In addition to the internal chargeinjection circuit the inputs of 16 cells of the firstcolumn are connected to an external pad It istherefore possible to connect the chip to a detec-tor in a simple way The drawback of such an ar-rangement is that it leads to a significant increaseof the input capacitance and therefore of the noise

All digital control signals are provided externallyin order to allow testing and optimisation of thespeed and synchronisation of the readout sequenceIt should be noted that the folding configurationis a slightly worse case with respect to the timingbecause of the longer bus lines


Test of the ASIC was performed with a DAQ chainbased on a NI-PCI (National Instruments) systemusing the PCI-7831R board and LabVIEW soft-ware The NI-FPGA generates pattern and worksas a triggered memory A Virtex II FPGA shiftsvoltage levels To allow the test of ToPix v2 witha 50 MHz clock a setup with a bidirectional FIFOhas been implemented on the FPGA board see fig-ure 324

Figure 324 Diagram of the data acquisition systemfor testing ToPix v2

The prototype has been tested both by injectingcharge through the inputs connected to the padsand through the internal calibration circuit Fig-ures 325 and 326 show the input-output transferfunction and the linearity over the whole dynamicrange for two typical pixels

Total dose irradiation tests have been performed onfour chips with a rate of 407 Gys Figures 327 and328 show the average of the baseline and the noiseover some of the 320 cells of 4 chips during irradi-ation obtained with a threshold scan in order toremove the quantisation error The baseline varia-tion is below 3 and is comparable with the disper-sion due to the process variation The noise level isaround 1 mV rms before irradiation which corre-


Figure 325 Input-output transfer function

Figure 326 Deviation from linear fit

Figure 327 Average baseline during irradiation

Figure 328 Average noise during irradiation

sponds to 0025 fC An increase of 20 to 003 fChas been observed after irradiation The measurednoise value if higher than the 0017 fC measured forthe first ToPix prototype It should be consideredthat in the ToPix v2 a large metal plate has beenconnected to the pixel input to emulate the bondingpad capacitance Simulations show that the capac-itance added by the metal plate justify the noiseincrease

After irradiation the chips have been annealed at100 C Figures 329 and 330 show a partial recov-ery in the baseline value and an almost completerecovery of the noise level to the pre-irradiation val-ues

Figure 329 Average baseline during annealing

Figure 330 Average noise during annealing

Figures 331 and 332 show the average and σ tailslope during irradiation In this case the variationis much more relevant up to 2 times in case of theaverage and up to 6 times for the σ The prob-lem lies in the very small current (around 6 nA)required to discharge the integrating capacitor toobtain the required resolution With such low cur-rents the leakage current induced by radiation playsan important role

A possible solution is to use an enclosed gate layoutfor the critical transistors of the discharge circuit


Figure 331 Average slope during irradiation

Figure 332 Tail slope sigma during irradiation

Such a layout technique has been proven to be effec-tive in preventing radiation-induced current leakage[15] However irradiation tests made on a previousprototype of the ToPix analogue part [13] showedthat at dose rate of 0011 Gys the relative ToTvariation decreases by an amount between 22 and54 in comparison with the variation after a doserate of 4 Gys

SEU Study

A dedicated study of the SEU in ToPix v2 has beenperformed in order to verify if the DICE architec-ture was sufficient in the expected PANDA environ-ment To this aim the device has been tested withheavy ions at SIRAD (LNL-INFN)[16] and the con-volution of the experimental results with simulationstudies allowed to estimate the maximum SEU rateof ToPix v2 in PANDA (see [17])More ions of different energies see table 38 havebeen used in order to have a wide range of depositedenergy Moreover it was possible to investigate thedeposited energy due to different incident angles

rotating the support of ToPix v2 [18] The irradi-ation runs lasted about 20 minutes and before andafter each run dosimetry measurements were carriedout

Ion Beam Angle Ecin [MeV] Edep [MeV]16O 0 101 07016O 30 101 08119F 0 111 09119F 30 111 10728Si 30 146 24035Cl 0 159 30158Ni 0 223 64779Br 0 215 896

Table 38 Ions used in the beam test with the cor-responding kinetic energy incident beam angle and de-posited energy in a sensitive volume of 1 microm3

The study of the configuration register was per-formed using a control program developed with theLabView software that wrote a 12 bit sequence ofalternated 0 and 1 every 2 seconds and countedwhenever some bits were changed Knowing thenumber of SEUs (Nerrors) and the total incidentparticle fluence (Φ) it is possible to calculate theprobability to have an upset usually called SEUcross section

σSEU =Nerrors

Φ middotNbit(31)

The σSEU has been calculated as the average of thecross sections coming from the several irradiationruns with the same ion species and the standarddeviation was used as an estimation of the errorThe experimental cross sections obtained have beenfitted by a Weibull function (see figure 333) It hasthe form

σ = σ0

[1minus eminus

(EdepminusE0

W

)s](32)

where W and s are shape parameters Here the sparameter was set to 3

Our fit parameters for the configuration register seefigure 333 are

bull Eth=04688 MeV

bull σ0 = 199 middot 10minus8 [cm2 bitminus1]

bull W = 4375 MeV

To evaluate the SEU rate of electronics developedin a standard layout without DICE architecture


Figure 333 Weibull fit to the experimental ion dataof the ToPix v2 configuration register A sensitive vol-ume of 1 microm3 is assumed

two other components of the end column circuit ofToPix v2 were tested for irradiationThe 12 bit shift register was tested using a methodsimilar to the previous one (for the configurationregister)The 12 bit counter was tested with a software com-paring at fixed time intervals its value with the oneof a ldquomirrorrdquo counter All the SEU tests were per-formed with a 10 MHz clock although the nominalclock of ToPix v2 is 50 MHz because of the longcable of the setupThe obtained fit parameters see table 39 in thesecases are very different from the Weibull param-eters obtained from the test of the configurationregister In particular the energy threshold is prac-tically zero this means that even a small depositedenergy can trigger an upset The saturation crosssection is bigger by a factor 2 in agreement withthe fact that a higher SEU rate is expected in elec-tronics designed in this standard layout

Device Eth [MeV] σ0 [cm2 bitminus1] W [MeV]

SR 158 middot 10minus15 26 middot 10minus8 103

CNT 731 middot 10minus14 26 middot 10minus8 146

Table 39 Fit parameters of the shift register (SR)and the 12b counter (CNT) in standard electronics

In order to obtain the SEU rate it is necessaryto know the probability for a certain particle todeposit a certain energy in the sensitive volume

Following the procedure explained in [19] therate of SEU for the PANDA hadron flux has beencalculated taking the lepton contribution to benegligible The evaluated SEU cross section Σin the hadron environment for the three testedelements of the ToPix v2 prototype has to becorrected with a 30 safety factor to take into ac-count the under-estimation of the sensitive volumeof 1 microm3 as explained in [19] The calculated Σsare reported in table 310

Σ [cm2bit]

CR SR CNT

1165 middot 10minus16 394 middot 10minus14 625 middot 10minus14

Table 310 SEU cross section in the hadron environ-ment corrected for the under-estimation of the SV con-figuration register (CR) shift register (SR) and counter(CNT)

Multiplying the Σ cross section by the hadron fluxobtained from a dedicated particle rate study [20]the SEU rates for annihilations on protons and onnuclei are obtained The results are reported intable 311

Annihilation CR SR CNT

pminus p 699 middot 10minus9 2364 middot 10minus8 375 middot 10minus8

pminus N 143 middot 10minus9 485 middot 10minus8 769 middot 10minus8

Table 311 SEU (bitminus1 sminus1) in the ToPix v2 proto-type configuration register (CR) shift register (SR)and counter (CNT)

Taking into account the final design of ToPix theSEU rate h can be derived taking into accountthat there are 12760 pixels the 8 bits for the config-uration registers and the 12 bits for the shift registerand the counter The estimated rates are listed intable 312 Considering the total number of chips(810) in the MVD pixel part we can estimate about2430 upsets per hour The SEU rates foreseen forthe shift register and the counter are higher in com-parison to the configuration register as expected


pminus p 3 130 207

pminus N 1 27 42

Table 312 SEU (chipminus1 hminus1) for the PANDA envi-ronment in the final ToPix prototype configurationregister (CR) shift register (SR) and counter (CNT)


The variation of one bit in the digital circuitry(SEU) following the passage of a particle can beproblematic in the PANDA environment The cor-ruption of the data is not a dangerous effect in itselfbecause the corrupted data can be easily identifiedand discarded Instead a more dangerous effect isthe loss of the detector control function that can berestored resetting or rewriting the chip As a conse-quence all the data stored from the upset event tothe reset operation have to be discarded As shownthere is a relevant different SEU rate between a ra-diation tolerant architecture such as the configu-ration register of ToPix v2 and the non-radiationtolerant architecture eg the shift register and thecounter Hence triple redundancy architecture ap-plication has been designed for the ToPix v3

ToPix Version 3

A third prototype of the ToPix chip designed towork with the 155 MHz clock updated to the newspecifications has been designed and tested Theprototype includes two 2times128 cell and two 2times32 celldouble columns The die size is 45 mmtimes 4 mmBump bonding pads have been used in the pixelcells for direct connections to the detector The128 cells columns have been folded in four 32-cellssub-columns in order to have an almost square formfactor thus simplifing the die handling The com-plete pixel cells and the end of column control logicand buffers have been implemented while the out-put multiplexer and serialiser is still in a simplifiedrevision The chip layout is shown in figure 334

Figure 334 ToPix v3 layout 45 mmtimes 4 mm dieThe pixel matrix is visible in the bottom part and theperipheral logic in the upper part Pads for wire con-nections are in the upper edge

In this prototype a clipping circuit has been imple-mented in order to avoid long dead-times for largeinput charges The circuit forces a ToT satura-tion at sim165 micros while keeping a good linearity forcharges up to 50 fCAll flip flops in the pixel cell are fully static and usethe triple redundancy technique in order to makethem more SEU tolerant The configuration reg-ister implements an asynchronous error detectionand self-correction circuitry while the leading andtrailing edge registers implement only the majorityvoter to save area Therefore the error correctionworks only in the readout phase In order to fit inthe pixel area the configuration register has beenreduced from 12 to 8 bitsThe chip is powered by a single 12 V power supplyDigital input and output drivers use the differentialSLVS standard [21] in order to reduce the digitalnoise contribution and to be compatible with the12 V power supplySimulations with layout parasitic extraction andprocess and mismatch variations show a gain of200022plusmn0046 nsfC at VTH = 30 mV and a chan-nel to channel ToT spread of 7 The simulatednoise level is 155 eminus at ILEAK = 0 and 209 eminus atILEAK = 10 nA

The Analogue Cell Layout is implemented in anarea of 50 micromtimes 100 microm The floorplan has beenoptimised in order to place the comparator be-tween the most sensible analogue blocks (preampli-fier constant current feedback and baseline holder)and the digital cell In figure 335 the different com-ponents of the analogue cell are shown

Charge Sensitive AmplifierThe bias lines are propagated vertically low levelsmetal are employed for the bias line and high lev-els metal for the power supply lines In order toavoid the propagation of spurious signals throughthe power supply 7 separate lines have been em-ployed The input cascode stage of the CSA has adedicated ground (gnd pre) The triple well tran-sistors of the input stage have a dedicated supplynet (vdd pre) as shield The leakage compensationstage bias is given by vdd ilc The comparator hasa dedicated supply line (vdd d) and a dedicatedground (gnd d) All the other analogue blocks aresupplied by vdd a and gnd a The input stage isa gain enhanced cascode amplifier with capacitivefeedback Cf A feedback capacitance of 12 fF whichis half of the one employed in ToPix v2 has beenchosen as the best compromise between the need ofmaximising the gain for low signals and the one ofhaving an adequate phase margin in the feed-backloop The output stage is a source follower which


Figure 335 Analog Cell Layout 1 Calibration cir-cuit 2 Charge Sensitive Amplifier 3 Constant Cur-rent Feedback 4 Baseline Holder 5 5 bit DAC for thethreshold correction 6 Comparator 7 Large chargecut-off circuit The octagonal shaped bump bondingpad area is shown in green

is employed as a voltage buffer to drive the outputcapacitive load The schematic view of this stage isshown in figure 336

The chip will be produced in a p-substrate processtherefore the standard NMOS transistors are fab-ricated directly on the chip substrate The pro-cess offers also the possibility of implementing theNMOS devices in dedicated wells (triple well op-tion) as sketched in figure 337 In this case a deepn-well is fabricated in which the p-well hosting theNMOS transistors is put This solution allows agood shielding of the transistor from the commonchip substrate at the expense of an increased area

Figure 336 Charge Sensitive Amplifier schematic

Figure 337 Triple well graphical representation (notto scale)

In a mixed mode design it is very important to pro-tect the sensitive nodes from the potential inter-ference of the digital switching circuitry The triplewell solution has hence been used for the input tran-sistor M1 and its cascode device M2

The CSA output stage is a source follower whichis employed as a voltage buffer to drive the loadingcapacitances at the input of the feedback stagesand of the comparator Since the source followerhas asymmetric rise and fall times and the polarityof the injected charge determines the direction ofthe leading edge a NMOS source follower is neededfor n-type sensors and a PMOS one for p-typesensors hence maximising the output swing Theselection between the two is done via the select polstatus bit In p-type configuration select pol islow M9 is switched on and M12 is switched offIn n-type configuration select pol is high M12 isswitched on and M9 is switched off This stage isbiased with a nominal current of 2 microA The outputDC voltage is regulated by the baseline holderThe value of the baseline voltage V ref must be setaccording to the sensor type Typical values are300 mV for n-type sensors and 700 mV for p-type


ones The buffers have the source and bulk shortcircuited in order to suppress the bulk effect thatwould decrease their voltage gain For this reasontransistor M11 in figure 336 is implemented as atriple well device in order to decouple its bulk fromthe global substrate and to allow the source-bulkconnection

Baseline HolderThe baseline holder is a specific feedback networkthat controls the DC output of the preamplifiermaking it insensitive to the sensor leakage currentAs a first step the solution presented in [22] has beenconsidered for the leakage compensation A similarcircuit has been designed in the current technol-ogy with a filtering capacitance of 10 pF in orderto evaluate its leakage compensation capabilitiesThis filtering capacitance requires an area of about500 microm2 which is close to the maximum area thatcan be allocated for that component The simula-tion results gave a ToT compression of 13 for a10 nA leakage current and of 38 for a 50 nA cur-rent Therefore a different approach based on adual feedback loop has been preferred This novelapproach shown in figure 338 is based on the factthat in deep submicron technologies MOS transis-tors with VGS = 0 can provide a resistance in theGΩ range Thanks to the high value of the equiv-alent resistance a capacitance in the order of 5 pFis sufficient to achieve a filter cut off frequency inthe order of 10 Hz The filtering capacitor is imple-mented through MOS devices exploiting the gateoxide and it occupies an area of 172 microm2

Figure 338 CSA with a feedback circuitry which au-tomatically compensates for the detector leakage cur-rent proposed in [22]

The output of the CSA is sensed by a differentialpair and compared to the preset reference voltageThe resulting error signal is used to drive a PMOSwhich acts as a voltage controlled current sourceand injects the compensation current into the inputnode

Constant Feedback Current GeneratorThe constant current feedback generator is a net-work that provides the CSA discharging current[23] Figure 339 shows the schematic view of thisstage The CSA output signal is presented to thegate of M2 while M1 receives the baseline referencevoltage With a p-type sensor Vout = 0 will have anegative polarity steering all the current from M1to M2 In this case the current in M3 is used torecharge the feed-back capacitor With n-type sen-sors Vout = 0 will be positive steering the currentfrom M2 to M1 M1 will have an excess current thatwill discharge the feedback capacitor At the equi-librium this stage provides an equivalent small sig-nal feed-back resistance of 67 MΩ with IFB = 5 nA

Figure 339 Schematic of the constant feedback cur-rent generator

This stage provides also the injection of the leakagecompensation current at the input node Usingfour switches it is possible to change the topologyof the circuit in order to choose the polarity of thecurrent to be compensated

ToT Linear Range LimitationWhen the CSA saturates the input node rises(p-type case) or drops (n-type case) In bothcases when the transistors connected at the input


Figure 340 Comparator schematic

node are pushed out from the saturation regionthe discharging current increases leading to acompression of the ToT signal This phenomenondepends on the input capacitance and the inputDC voltage Since the input DC voltage is atsim 200 mV assuming a detector capacitance of200 fF for an n-type sensor the compression occursfrom Qin asymp 90 fC For a p-type sensor a muchlarger current (sim 200 fC) could be collected on theinput capacitance while ensuring the proper biasof the constant current feedback

ComparatorThe comparator has a folded cascode input stageand two CMOS inverters in cascade as driver inorder to have a fast transition between the twologic states Since the comparator performancein ToPix v2 was satisfactory the transistor levelimplementation has not been changed Howeverthe layout has been redesigned in order to fit in theavailable cell area Figure 340 shows the transistorlevel implementation The comparator gives ahigh level output when Vin2 gt Vin1 Thereforein n-type configuration where the CSA outputis positive in1 is connected to the comparatorreference voltage and the CSA output to in2 Inp-type configuration where the CSA output isnegative the signals are exchanged using CMOSswitches

DAC Current SourceSince Vref d is common to all pixels in order tomitigate the threshold dispersion a local 5 bit DACis added in each pixel to allow a fine tuning of thethreshold on a pixel by pixel basis

A simplified scheme of the DAC is shown infigure 341 For simplicity the switches are repre-sented in common for each group of Ibit current

Figure 341 Simplified model of the 5 bit DAC

generators Each of the bit B0 B1 B2 B3controls a switch that activates the relative currentgenerator

Clipping CircuitA clipping circuit has been designed in orderto protect the front-end when a large charge ispresented at the input node thus avoiding longdead times The front-end requires to measurecharges up to 50 fC so larger charges have to bedischarged with an extra current The clippingcircuit exploits the fact that the input node rises(p-type case) when the CSA saturates Connectingthe input node at the gate and drain of a low powerNMOS and its source at ground it is possible toclip the voltage signals larger than the thresholdvoltage of the transistor (Vth asymp 600 mV)

Calibration CircuitThe calibration circuit allows to inject a currentpulse (Iinj) into the preamplifier to test the circuitresponse Figure 342 shows the schematic of thecalibration circuit The TestP signal is a CMOS dif-ferential signal common to all the pixels The posi-tive line is TestPH and the negative one is TestPLTestP EN is a status bit of the pixel configurationregister and activates the injection subcircuit

Figure 342 Calibration circuit schematic



The ToPix v3 prototype has been tested on a benchto verify the performances The experimental setupforesees the testing board housing ToPix v3 and aXilinx evaluation board equipped with the Virtex6 FPGA The acquisition system is based on theLabView software Due to a problem in the col-umn buses only the 32-cells column can work at160 MHz therefore all the tests have been per-formed at 50 MHz The problem has been inves-tigated and is partially due to the folded structureof the column which makes the bus significantlylonger (by a factor of 13) and partially from thefact that the 3 latches of each bit (due to the tripleredundancy) are directly connected to the bus with-out any buffer thus tripling the capacitive loadFigure 343 shows the preamplifier output for dif-feret input charge values obtained with a thresholdscan and the internal test signal generator Theobtained shapes show the fast rising edge of thecharge amplifier and the slow discharge at constantcurrent even in the saturation case

Figure 343 Reconstructed signal shapes with thresh-old scan for different input charge values

At present four ToPix chips have been tested show-ing similar performances Figure 346 shows thepixel response of the four tested samples In thefirst row the chips which respond correctly to a se-quence of 4 test pulses are shown in green Pixelstransmitting twice the correct number are markedin white while pixels which do not respond at allare in black Intermediate numbers of responsesare shown accordingly with the colour scale Inthe second and third rows the leading and trail-ing edge values are represented respectively Thetime stamp values are represented in a colour scaleAll the test signals are sent at the same time andwith the same charge therefore a uniform colourpattern is expected Here pixels in black and in

white correspond to a all 0rsquos and all 1rsquos output pat-tern respectively It can be seen that the number ofdefective pixels is slightly higher apart from chip 2where a column shows significant problems It canalso be observed that the variation of the trailingedge value is higher than the leading edge Thisvariation is expected due to the higher variation inthe charge injection system and to the higher noiseof the trailing edge

Figure 344 shows the baseline level for all 640 pixelsof a chip before and after baseline equalisation

Figure 344 ToPix v3 extrapolated baseline beforeand after equalisation

Figures 345 shows the baseline distribution before(left) and after (right) correction A baseline sigmaof 5 mV has been measured before correction whichcorrespond to an equivalent charge dispersion of473 eminus After the correction with the on-pixel 5 bitDACs a dispersion of 650 microV corresponding to lessthan 62 eminus can be obtained The measurementsconfirm the effectiveness of the correction DAC toreduce the threshold dispersion

Figure 345 ToPix v3 baseline distribution before andafter correction

Figure 347 shows the measured gain The mea-sured value of 66 mVfC is slightly lower than thesimulated one (75 mVfC) but compatible with theprocess variations and with the uncertainties of thecharge injection mechanism


Figure 346 ToPix v3 pixel map for the four tested samples

Figure 347 ToPix v3 measured analogue gain

Figure 348 shows the measured noise The averagevalue of 100 eminus obtained without the detector con-nected to the preamplifier input is consistent witha simulated value of 150 eminus with the detector ca-pacitance included

Figure 348 ToPix v3 measured noise

The charge to time overall transfer function isshown in figure 349 for the full charge input rangeand in figure 350 for charges up to 6 fC Also shownin figure 349 is the deviation from the linear fit

Figure 349 ToPix v3 transfer function for nominalfeedback current - full range

Due to the reduced clock frequency the chip trans-fer function has been measured also for a feedbackcurrent which is half of the nominal value in orderto partially compensate for the reduction in chargeresolution The results are shown in figure 351 forthe full range and in figure 352 for the first part ofthe dynamic range


Figure 350 ToPix v3 transfer function for nominalfeedback current - 06 fC range

Figure 351 ToPix v3 transfer function for halvedfeedback current - full range

34 Hybridisation

Both the electrical and mechanical connections be-tween the sensor and the readout chip are made bysolder balls in a flipndashchip solderndashbonding processThis technique is now well developed for the appli-cation in the PANDA experiment not only for thedaily connection process established in the FPGAproduction but also for the strong experience ac-quired in the LHC experiment realisation comparedto the first solder flipndashchip process developed byIBM in 1969 [24]

341 Bump Bonding

Many companies (IZM VTT SELEX) have the ca-pacity to perform the bump bonding process Sol-

Figure 352 ToPix v3 transfer function for halvedfeedback current - 06 fC range

der balls made of SnndashPb or In have been used withgood results in the LHC experiments Balls sizes upto 20 microm have been deposited on pads correspond-ing on one side to the pixel sensor and on the othersize to the single readout cell This technique canbe regarded as well known Pads on new epitax-ial material and circuits made by the new 130 nmCMOS technology have been checked The test ofthe first one is reported in the following subsectionand test of the second one has been performed inan other experiment [25]

First Thinned Prototypes

With the first thinned prototypes described at thebeginning of Section 32 the quality of the bump-bonding process was measured using a 90Sr sourcein self-triggered mode Figure 353 shows the num-ber of dead pixels for the tested assemblies (8192pixels per assembly) [3] In this case the termldquodeadrdquo is used to cover several possible reasons fora pixel not being usable for signal acquisition elec-tronics channels with zero-gain high channel noisedue to high individual pixel leakage current andfailed bump connection Apart from the S6 assem-bly which had an obvious localised problem coveringa region of pixels the connection yield and overallefficiency is very high better than 99 An exam-ple of the source profile for the S8 (Epi-75) assemblyis shown in figure 354

During the realisation of the first thinned sensorunfortunately all the epi wafers broke in the finalstages of the thinning process either during clean-


Figure 353 Dead pixel counts in the tested assem-blies

Figure 354 Source profile obtained from S8 (Epi-75)assembly using a 90Sr source the x-axis correspondingto the pixel column number (32 in the ALICE readout)and the y-axis to the row number (256 in the ALICEreadout)

ing or during removal from the protective tape usedto hold the wafers during thinning One possible ex-planation is stress caused by significant differencesin the lattice constants of the high-purity epitax-ial layer and the heavily doped substrate Thendevelopment work in collaboration with the VTTcompany was concentrated on understanding thisproblem better with the aim of resolving it

In particular a set of 12 patterned wafers featur-ing 50 75 and 100 microm epitaxial thickness with de-posited bumps were thinned on the Cz side up tothe target thickness of 200 microm This final thicknesswas selected with the aim to evaluate a possiblecontribution to the wafer breakage by the last pas-sivation processes on the top of the patterned waferoptimised for 200 microm thickness as in the ALICE ap-plicationOnly two wafers were destroyed during thethinning process giving a yield of 83

Then 9 blank wafers with 100 microm epitaxial layerwere thinned in groups of three to the final thick-nesses of 130 120 and 110 microm by removing most ofthe Cz substrate The first two wafers broke during

the CMP process used to remove the defect layersimmediately after the backgrinding The etchingtechnique was planned at this time and no otherbreakage was detected with this last technique

35 Single Chip AssemblyPrototype

The first prototype of single chip assembly made ofan ASIC developed in 130 nm CMOS technologyToPix v3 and thinned epitaxial silicon sensor willbe produced as the final step of the RampD phase forthe hybrid pixel detector development

The assembly is based on ToPix3 already describedin the readout prototype section The reduced scaleof the pixel sensor matrix asks for a dedicated sensorlayout Figure 355 shows the pixel matrix layoutmade of 640 100 micromtimes 100 microm pixels to overlap thereadout cell matrix of ToPix v3 On the perimeterthe guard ring sequence the first one connected toground and the others in a floating configuration isvisible

Figure 355 Pixel matrix layout

A second layout has been foreseen taking into ac-count the larger pixels needed across the sensorregion where two readout chips will be arrangedside by side in the final project This last proto-type is made of 600 100 micromtimes 100 microm pixels and 40100 micromtimes 300 microm pixels

Figure 356 shows the schematic assembly betweenToPix v3 and the sensor

The pixel sensors have been arranged in 4 inch epi-taxial wafers and produced at FBK In particularMR and LR wafers as described in the RadiationDamage section have been used Besides a wafer ofeach type was oxygen enriched Figure 357 showsa sensor view by means of a microscope


Figure 356 Scheme of the assembly The sensor (inblack) is placed above the chip (in gray)

Figure 357 Photograph of epitaxial sensor producedat FBK Pixel matrix matching ToPix v3

36 Module

The philosophy behind the design process of thepixel modules is to first optimise forward coveragethen adjust barrel design as a result Only one vari-able is to take into account to perform this studythe chip size Nevertheless a small production ofsilicon sensors as foreseen must fit on 4 inch wafers

A hybrid pixel module is composed of several read-out chips bump bonded to a unique silicon sensoron the top of the sensor a multilayer bus routingboth power supply lines and signal lines is foreseenequipped with minimal SMD components Bus andreadout chips are connected by wire bonding tech-nique Potentially a chip controller of this modulecould be arranged on the top of the bus to driveGbit links and serving two or three readout chips

Starting from the disks the most useful sensor lay-

out appears to be an alternating front and backconfiguration with respect to the carbon foam layerused to increase the heat transport to the coolingpipes embedded in the middle of this layer and act-ing as a mechanical support too In particular apixel active area of 116 mmtimes 114 mm per chip isforeseen and 2 4 5 6 readout chips sensors (S2S4 S5 S6 respectively) are planned with verti-cal gaps bridged with longer pixels as already ex-plained in 32 The nominal readout chip size is148 mmtimes 114 mm or slightly less to include dic-ing tolerances Figure 358 shows an S2 modulescheme

Figure 358 S2 module two readout chips are bumpbonded to a single sensor

The small disks are equipped with six S2 modulesand two S4 modules with horizontal gap and asstarting point a sensor overlap due to the dead spacedriven by the conventional guard ring design Ageometrical total coverage of 65 is reached buta systematic small shift will be studied to investi-gate the spatial distribution of the emitted particletracks to maximise the reconstructed hits Fig-ure 359 shows the module configuration of a smalldisk

Figure 359 Module configuration of a small diskThe radius is 3656 mm

The big disks will be equipped with four S2 four S4


twelve S5 and four S6 modules In this case a smalldead zone (of about 1 mm) will result where twosensors meet in the horizontal plane Figure 360shows the module configuration of a big disk

Figure 360 Module configuration of a big disk Theradius is 7396 mm

In general keeping cables out of the active regionwill mean that some modules may require two de-signs according to which end the cables have to beconnected

In figure 361 the complete forward assembly isshown

Figure 361 The complete forward pixel assemblyscheme

The pixel barrels are made of two layers of staveswith modules featuring differently sized sensors toleave out the target pipe The first layer houses8 staves made of S4 modules and six staves basedon S2 module each The second layer is assembledfrom six staves made of S5 modules each and 22staves housing two S6 modules each

37 Bus-Flex Hybrid

The bus on the top of the sensor in a hybrid pixelmodule routes electrical signals and power linesFor minimising material and for handling high datatransmission different approaches in particular forthe electrical signals have been followed studyingthis part of the project taking into account the twodifferent solutions with and without a module con-troller serving two or three readout chips

A first approach is based on a custom solutionwhere at least two aluminium metal layers are in-terconnected with vias The major parameters arealuminium layer thickness of about 5 microm a mini-mum track pitch le 100 microm a minimum via diam-eter le 25 microm lowest possible dielectric constant tominimise thickness with the impedance control thetop layer compatible with wire bonding and SMDmounting at least the overall structure size has tobe compatible with a 4rdquo Si wafer

In figure 362 different size buses useful for differentpixel modules are arranged inside an area suitablefor a 4rdquo Si wafer

Figure 362 Busses for different pixel modules

The bus layouts are designed for the case the mod-ule controller is not foreseen and a high density oftracks has to be handled and routed on the bus firstlayer up to a aluminium strip cable (see the Cablesection in the Infrastructure chapter 54) The sec-ond layer avoids the track crossing eg of the clockline in case a single line is used for the whole mod-ule

The feasibility of such a solution was investigated


with a first company (Techfab Italy) The base ideawas to build the bus structures on silicon wafer assupport and then take advantage of the Sindashdetectorprocesses and quality lithography SUndash8 was identi-fied as the most interesting dielectric for its relativedielectric constant (about 3) This lower value al-lows a dielectric thickness of about 15 microm betweenAl layers The electroless copper deposition wasproposed as an useful process for filling the vias(figure 363)

Figure 363 Scheme of the SU8 technique

Techfab went out of business after the very first pro-cess tests so that the project will now be pursued incooperation with FBK Trento First test of SUndash83010 was performed by using a strip-shaped maskA layer that was only 196 microm thin has been ob-tained as shown in the photograph of figure 364where the wall of the strip shows a reverse angle(figure 365) not perfectly suited for aluminium de-position

Figure 364 SU8 strips

Besides an electroless technique with nickel appearsto be better suited for the cleanliness of the manu-facturing equipments

In parallel to this new approach a base solutionhas been considered using techniques developed atCERN for the Alice experiment In this case amultilayer bus with a staircase shape (figure 366)

Figure 365 Slope of an SU8 strip

houses deposited aluminium tracks featuring 80ndash90 microm width and 150 microm pitch on each layer andmicrovias between two layers have been made

Figure 366 Staircase multilayer

Alternatively to 10 microm aluminium deposition15 microm aluminium foil glued to polyimide has beenconsidered Different etching processes have beenstudied for both solutions [26]

As an example in figure 367 a layout of a busforeseeing a module controller for the readout chipsis shown

The power supply routing bus can be made of alu-minium strips glued to polyimide using the sametechnique that will be explained in the Infrastruc-ture chapter Cable section (54) for data trans-mission Analogue and digital power supply linesplanned for the readout chips and the bias voltage


Figure 367 Bus with module controllers

for the sensor with corresponding ground planes willbe the first layers of the staircase bus on the top ofthe sensor

References

[1] G Kramberger et al Superior radiation toler-ance of thin epitaxial silicon detectors NuclInstr Meth A515665ndash670 2003

[2] E Fretwurst Recent advancements in the de-velopment of radiation hard semiconductor de-tectors for S-LHC RD50 Workshop Helsinki2005

[3] D Calvo et al Thinned epitaxial silicon hybridpixel sensors for the PANDA experiment NuclInstr Meth A594 2008

[4] F Antinori et al Infrared laser testing of AL-ICE silicon pixel detector assemblies Nucl In-str Meth A56813ndash17 2006

[5] F Antinori et al The ALICE pixel detectorreadout chip test system Proceedings of Sev-enth Workshop on Electronics for LHC exper-iments CERN-LHCC-2001-034 2001

[6] A Tengattini Radiation damage effects onepitaxial silicon devices for the PANDA exper-iment Masterrsquos thesis Universita degli Studidi Torino 2010

[7] G Lindstrom et al Proceedings of the 10thEuropean Symposium on Semiconductor De-tectors Nucl Instr Meth A568-11ndash4742006

[8] F Faccio and G Cervelli Radiation-inducededge effects in deep submicron CMOS transis-tors IEEE Trans Nucl Sci 522413ndash24202005

[9] I Peric et al The FEI3 readout chip for theATLAS pixel detector Nucl Instr MethA565178ndash187 2006

[10] R Beccherle et al MCC the module controllerchip for the ATLAS pixel detector Nucl InstrMeth A492 2002

[11] S Schnetzer for the CMS Pixel CollaborationThe CMS pixel detector Nucl Instr MethA501 2003

[12] A Kluge et al The ALICE silicon pixel detec-tor Nucl Instr Meth A501 2007

[13] D Calvo et al ToPix The first prototype ofpixel readout for PANDA experiment NuclInstr Meth A59696ndash99 2008

[14] T Calin M Nicolaidis and R Velazco Upsethardened memory design for submicron CMOStechnology IEEE Trans Nucl Sci 432874ndash2878 1996

[15] G Anelli et al Radiation tolerant VLSI cir-cuits in standard deep submicron CMOS tech-nologies for the LHC experiments Practi-cal design aspects IEEE Trans Nucl Sci461690ndash1696 1999

[16] D Bisello et al The SIRAD irradiation facilityfor bulk damage and single event effects stud-ies Proceedings of the 2nd SIRAD WorkshopLNL INFN pages 426ndash434 2004

[17] L Zotti et al Study of the single event upseton ToPix 2 PANDA MVD-note 008 2011

[18] LW Connell et al Modeling the heavy ionupset cross section IEEE Tran Nucl Science42 1995

[19] M Huhtinen and F Faccio Computationalmethod to estimate single event upset rates inan accelerator enviroment Nucl Intstr MethA450155ndash172 2000

[20] L Zotti D Calvo and R Kliemt Rate Studyin the Pixel Part of the MVD PANDA MVD-note 007 2011

[21] JESD8-13 JEDEC Scalable Low-Voltage Sig-naling for 400 mV (SLVS-400)


[22] F Krummenacher Pixel detectors with lo-cal intelligence an IC designer point of viewNucl Instr Meth A305(3)527ndash532 1991

[23] I Peric Design and realization of integratedcircuits for the readout of pixel sensors in high-energy physics and biomedical imaging PhDthesis Universitat Bonn 2004

[24] PA Totta and RP Sopher SLT device met-allurgy and its monolithic extension IBM ResDevelop 13 N3 1969

[25] M Noy et al Characterization of the NA62Gigatracker and of column readout ASICTWEPP 2010

[26] Rui de Oliveira Low mass ALICE pixel busALICE Workshop 2010


67

4 Silicon Strip Part

This chapter describes the silicon strip part of theMVD which consists of a barrel and a forward disksection as outlined in section 24 The basic designforesees double-sided strip detectors of rectangularand trapezoidal shape respectively The overallconcept of the MVD is based on a standard processfor radiation hard sensors with a target thicknessof 200 to 300 microm It follows approved solutions ofother tracking systems already installed at HEP fa-cilities As for the entire apparatus the trigger-lessreadout concept represents one of the major tech-nical challenges and requires the implementation ofnew technologies Moreover the readout on bothsensor sides requires sophisticated technical solu-tions for the hybridisation of the detector modules

41 Double-Sided Silicon StripDetectors (DSSD)

Presently silicon strip detectors are deployed to agreat extent as tracking detectors in particle physicsexperiment worldwide The division of the semi-conducting detector by means of strips enables aone-dimensional spatial resolution of particle trackstraversing the detector The utilisation of double-sided strip detectors offers the ability of precisepoint reconstruction with the advantage of less de-tector channels compared to pixel detectors How-ever ambiguities occur in case of multiple parti-cles crossing the detector in the same time frameTherefore the PANDA MVD utilises silicon strip de-tectors at the outer layers of the MVD where theparticle flux is reduced in contrast to the inner lay-ers

411 Barrel Sensors

First PANDA full size prototype sensors have beenproduced in the first half of 2011 by the companyCiS GmbH in Erfurt (Germany) The productionprocess utilises a 10+2-layer process with a sin-gle metal layer laid out on 4primeprime-wafers consisting oflang111rang - cut material A floorplan of the designedlayout of the implemented structures can be seen infigure 41

Table 41 gives an overview of the specifications ofthe wafer and of the implemented sensors As start-ing material a monocrystalline Floating Zone (FZ)

Figure 41 Floorplan of wafer with full size prototypesensors The Sensor properties are comprehended intable 41

silicon wafer was chosen For the implemented sen-sor structures front side strips are realised as p+

in n doping whereas the back side contains n+-strips embedded in the n-substrate All sensors aredouble-sided with strip structures oriented orthog-onal to each other on either side The depletion isachieved by punch-through biassing from the biasring towards the strips on both front and back sideand the n-side charge separation is realised by p-spray implants The active area is protected byeight guard rings to assure a stable electric fieldwithin the active area In this scheme front sideis considered the junction contact or p-side whilethe ohmic or n-side is always referred to as backside

Besides the full size PANDA prototype sensors asmaller square shaped sensor S3 with 50 microm pitchand an active area of roughly 2 times 2 cm2 was realisedon the delivered wafer (see figure 42) This sensorwas foreseen to be an equally sized replacement foralready existing prototype sensors from earlier pro-totyping stages in order to benefit from the existingreadout infrastructure Furthermore five ldquoBabyrdquo-Sensors (S4) with only 128 strips on either side wereplaced on the wafer Additional test structures areimplemented as well serving mainly as markingsduring wet-processing stages bonding calibrationtags or similar Finally fourteen diodes with differ-ent numbers of guard rings are placed around the


General

wafer material FZ Si 4primeprime nP

thickness 285 plusmn 10 microm

resistivity 23 50 kΩ middot cm

n-side isolation p-spray

guard rings 8

stereo angle 90degpassive rim 860 microm

S1

n-side strips 896

p-side strips 512

pitch 65 microm

active area 58275 times 33315 mm2

S2

n-side strips 512

p-side strips 512

pitch 65 microm


S3

n-side strips 384

p-side strips 384

pitch 50 microm


S4

n-side strips 128

p-side strips 128

pitch 65 microm


Table 41 Specifications of the prototype strip sensors

Figure 42 4primeprime-Wafer with prototype sensors

main elements on the wafer These diodes may beused to derive the radiation dose during irradiationtests

In figure 43 a corner of the sensors on the n-side(left picture) and one on the p-side (right picture)

Figure 43 Layout detail of MVD-Barrel Double-Sided Strip Sensors ohmic- or n-side (left) junction-or p-side (right)

are shown with the first strips and the high poten-tial bias contact ring with square-shaped passiva-tion openings The pads on the n-side strips areconnected to the odd-numbered n+-implants whilethe even-numbered pads are located at the oppo-site edge (not shown) On the junction- or p-sideadditional guard ring structures for insulation of thehigh potential difference to the bias line are neces-sary The contacts along the inner bias line are usedfor the supply of the negative bias potential TheDC-pads on the strips are direct connections to thep-side p+-implants For ease of channel identifica-tion additional orientation numberings and posi-tion marks are implanted inside the metal layer

A measurement of the leakage current flowingthrough the sensor with applied variable reversevoltage at the bias contacts is shown in figure 44The recorded curves are the I-V - characteristics ofthe large sensors (60 times 35 cm2) from 25 differ-ent wafers out of two lots Almost all of the sen-sors show the expected behaviour ie very low andconstant leakage current above full depletion anda steeply increasing current at reverse breakdownvoltage Two of the analysed sensors obviouslycompletely fall off this pattern one which showsa continuous strong rise of the current even at lowvoltages and one that appears to become depletedat all but shows excessive leakage at higher voltagesIt is already clear from this first picture that sen-sors past production should be casted into qualitycategories to ensure equal parameters of the sensorsutilised in a collective context (in this case the mod-ules inside the MVD) The full depletion voltage foreach sensor is extracted before the actual assemblytakes place This is done in a probe station setupby sampling the capacitance vs bias voltage char-acteristics as seen in figure 45 The p-side singlestrip capacitances are measured with typical valuesof asymp 450 fF at a depletion voltage of 150 V For com-parison the total capacitance at the bias contacts(black trace) is shown In order to assure error-free


0

5

10

15

20

0 50 100 150 200 250 300 350

Leak

age

Cur

rent

microA

Bias Voltage V

IV-Characterization Lots 310344+310487 Sensor Type T1

01020406070809101112131415

161718192022232425262728

Figure 44 I-V-Curves of full size (60times 35 cm2)silicon sensor prototypes originating from 25 differentwafers At full depletion the majority of the sen-sors shows a nearly constant leakage current betweenasymp 1 4 microA over a wide reverse voltage range Thesteep increase at high voltages is due to the avalanchebreakdown and varies significantly from wafer to waferThe temperature during the measurements was heldconstant at 20 C

operation inside the PANDA-detector a procedure issuggested that allows the selection of sensors whoseparameters satisfy following criteria

bull the global leakage current must not exceed10 microA below breakdown voltage

bull the global capacitance must show clear deple-tion above full depletion voltage

bull the breakdown voltage must be at least 50 Vhigher than the maximum foreseen bias voltageof 200 V

In the first 2011 prototyping batch 80 of the deliv-ered sensors (type S1) conformed to these criteriaIt is expected that the yield increases once the pro-duction parameters are fixed and the producing siteimposes post production tests allowing the deliveryof sensors according to these criteria [1]

For the qualification of radiation hardness the S4sensors (128 strips on either side with a pitch of65 microm) have been irradiated with protons andneutrons of several fluences in order to categorisethe obtained sensors with respect to the system-atic radiation damage studies undertaken by theRD48 collaboration [2] Particularly the definitionof the maximum depletion voltage of 200 V nec-essary to operate irradiated sensors at the end ofthe full PANDA lifetime with an applied fluence of1middot1014 n1MeV eq cmminus2 is derived from these collected

Bias Voltage [V]0 50 100 150 200

Cap

acit

ance

[p

F]

0

5

10

15

20

25

30

35

whole Sensor

strip 21

strip 27

strip 524

strip 736

strip 740

Figure 45 Capacitance characteristics of p-side stripsof one large (60times 35 cm2) prototype sensor At fulldepletion the capacitance of single strips lies well below1 pF The capacitance between bias and bulk contacts(black) is shown for comparison

Figure 46 Effective bulk doping concentration andrequired depletion voltage vs 1 MeV equivalent neutronfluence for standard Floating Zone and oxygen enrichedsilicon sensors [2]

data (see figure 46) With this choice an addi-tional oxygen enrichment stage is not mandatorybut left as optional degree of freedom to compen-sate for possible yield losses Irradiations with slow(14 MeV) protons were carried out at the cyclotronfacility in Bonn with a total fluence of up to 22middot1013

pmiddotcm-2 which corresponds to a 1 MeV neutronequivalent fluence of roughly 8 middot 1013 cm-2 obtainedby NIEL scaling I-V and C-V - characteristics ofthe irradiated sensors (type S4) have been measuredin different time intervals after the end of the irradi-ation (figures 47 and 48) The measurements tookplace between annealing intervals of 20 hours (redcurves) 310 hours (green curves) and after an an-nealing at a temperature of 60 C for 24 hours (bluecurves) The green curves correspond to the stan-dard annealing interval of 80 min in a 60 C envi-


Bias Voltage [V]0 20 40 60 80 100 120

Lea

kag

e C

urr

ent

[uA

cm

^3]

1

10

210

310

before irradiation

20h

310h

5100h

Figure 47 Annealing behaviour of leakage cur-rent characteristics after irradiation of S4 sensors with10 MRad 14 MeV protons The I-V - trends of thesensors were recorded after 20 hours (red) 310 hours(green) and 5100 hours (blue) of annealing time nor-malised to 25 C

Bias Voltage [V]0 20 40 60 80 100 120

Cap

acit

ance

[p

F]

3

10before irradiation

20h

310h

5100h

Figure 48 Annealing behaviour of capacitance char-acteristics after irradiation of S4 sensors with the an-nealing times given in figure 47

ronment as proposed by [2] The I-V - trends in fig-ure 47 are normalised to the volume leakage currentin order to have a proper comparison to measure-ments available from other groups The obtainedleakage current of asymp 1 mA middot cm-3 after irradiationmeets well with the value to be expected from fieldstudies [3] The recorded post-irradiation C-V -trends in figure 48 show nicely the expected bene-ficial (red and green) and reverse (blue) annealingbehaviour as well as the change in the full deple-tion voltage from 60 V to 50 V due to the change inthe doping concentration according to figure 46 andthus demonstrate the suitability of the first batchof prototype sensors for the targeted requirements

412 Wedge Sensors

Basically the technological option chosen for thewedge sensors will be the same than for the bar-rel sensors A picture of the main sensor layout isshown in figure 49

Figure 49 Dimensions of the strip wedge sensor

The active area of 1517 mm2 stays below the onefor the rectangular barrel sensor (1941 mm2) so thatmajor characteristics of the sensors are expected tobe in the same range than the ones of the alreadytested full size barrel prototypes In contrast to thebarrel sensor the readout pads on both sides willbe at the same sensor edge Strips run parallel toone of the edges of the trapezoid Thus those onthe opposite side get shorter when approaching thesensor edge A connection scheme adapted from [4]is shown in figure 410 Basic design parameters ofthe sensor are listed in table 42

Figure 410 Illustration of the pad geometry at thetop edge of the sensor (adapted from [4])


Strip pitch 675 microm

Strip orientation to sensor edge

Number of strips 512 side

Stereo angle 15

Sensor height 5767 mm

Total area 1688 mm2

(Active area) (899)

Table 42 Design parameters of the wedge sensor

42 Front-End Electronics ASIC

421 Requirements

The data acquisition concept of the PANDA detec-tor demands front-end electronics that is able to runwithout the need of an external trigger Thereforethe front-end must be able to distinguish detectornoise from physical events and send this hit infor-mation together with a precise time stamp to thedata acquisition

A decision towards the front-end to be deployed forthe strip part of the PANDA MVD has not yet beenmade The front-end should feature a measure-ment of the deposited charge of a particle cross-ing the detector in order to achieve a spatial reso-lution through clustering algorithms This energyloss measurement may also be used for a particleidentification hypothesis (PID) Since only digitalinformation about the charge measurement shouldbe transfered from module level the digitisationshould be done on the front-end Digitisation canbe achieved via the Time-over-Threshold method(ToT) or sampling-ADCs Regardless whether ToTor ADCs will be utilised a low power consump-tion should be aspired The digitisation resolutionshould be in the order of 8 bit or better to ensure aprecise charge measurement and moreover a highspatial resolution Assuming linearity of the digi-tiser within a dynamical range of 10 minimum ion-ising particles 8 bit are required to assure that theleast significant bit will be comparable with the onesigma detector noise In order to save board spaceand number of data cables the front-end should fea-ture highspeed serial data links The front-end shallbe radiation hard to a level of up to 10 MRad Thetime-stamping of observed hits will be done withthe PANDA timing clock of 155 MHz Table 43summarises the requirements for the strip readoutfront-end

422 Options

Currently there are no front-end ASICs availableto fully comply with the demanded requirementsHowever there are developments for self-triggeringfront-ends which are mostly in an early prototypingstate The available and future developments willhave to be tested with respect to the demands of theMVD Some of the front-end options are presentedbelow

Modified ToPix Version 3

The use of very similar front-end architectures forboth pixel and strip sensors in the MVD opens in-teresting perspectives In this case in fact theback-end electronics for the two sub-systems couldshare the same hardware design with minimal cus-tomisation being done at the firmware and soft-ware level Such an approach would allow impor-tant synergies inside the MVD community reduc-ing the development time and costs of the overallproject However the two sub-detectors differ sig-nificantly in key specifications such as the sensorcapacitance (few hundreds of femtofarad for the pix-els and several picofarad for the strips) data rate(few kilohertz per pixel versus about ten kilohertzper strip) and power consumption per channel (fewmicrowatts in case of pixels and few milliwatts forthe strips) From these numbers it is apparent thatthe two parts of the MVD demand a detailed cus-tomisation of the very front-end electronics

In ToPix the front-end ASIC for the pixel sensorsdescribed in chapter 333 the Time over Thresh-old technique (ToT) has been adopted to digitisethe charge information In this approach the feed-back capacitor of a high gain amplifier is dischargedwith a constant current The time during which theamplifierrsquos output remains above a given thresholdis measured by counting the clock pulses One ofthe interesting features of the method is that thelinear dynamic range is extended well beyond thesaturation point of the amplifier This combineslow power operation with the capability of measur-ing energy losses much greater than those ones ofminimum ionising particles

In the present design the input amplifier employsa single stage topology This choice was dic-tated by the limited space available in the pixelcell for the implementation of the analogue part(50 microm times 100 microm) One of the drawbacks of thisconfiguration is the increased sensitivity to cross-talk for large input signals that saturate the am-plifier When this happens the amplifierrsquos open-


Parameter Value RemarksGeometry

width le 8 mm

depth le 8 mm

input pad pitch asymp 50 microm

pad configuration lateral pads occupied only for diagnostic functions shouldbe left unconnected for final setup

channels per front-end 26 28 default 128 channels

Input Compliance

sensor capacitances

fully depleted sensor

lt 10 pF rect short strips

lt 50 pF rect long strips + ganging

lt 20 pF trapezoidal sensor

input polarity either selectable via slow control

input ENC lt 800 eminus Csensor = 10 pF

lt 1000 eminus Csensor = 25 pF

Signal

dynamic range 240 keminus (asymp 385 fC)

min SNR for MIPS 12 22500 eminus for MIPs in 300 microm sil-icon guaranteed within lifetime

peaking time asymp 5 25 ns typical Si drift times

digitisation resolution ge 8 bit

Power

overall power dissipation lt 1 W assuming 128 channels per front-end

Dynamical

trigger internally generated when charge pulse exceeds ad-justable threshold value

time stamp resolution lt 20 ns

dead time ch lt 6 micros baseline restored to 1

overshoot recovery time ch lt 25 micros

average hit rates ch derived from simulations at

(poissonian mean) a beam momentum of 15 GeVc

hot spots 9000 sminus1 pp

40000 sminus1 pAu

average occupancy 6000 sminus1 pp

30000 sminus1 pAu

Interface

slow control any low pincount eg I2C

data sparsified data

Table 43 Requirements for the strip front-end ASIC

loop gain drops thereby determining an enhance-ment of the cross-talk between adjacent channelsthrough the inter-electrode capacitance For pixelsensors this capacitance is small and such an effectcan be tolerated However the saturation of the in-put stage would be a serious issue for strip sensorswhere the inter-strip capacitance may reach severalpicofarads

The measure of the charge with the ToT gives avariable dead-time The high granularity of the pix-els (cell size of 100 microm times 100 microm) reduces the eventrate per channel to the kHz level and dead timesof several microseconds can hence be tolerated Forstrip sensors on the other hand the event rate perchannel is significantly higher


-+

-+

-+

-+

ToT stages (time stamp and charge measure) Output logic

Current buer

Cf

Cz

Rz

Rf

Cd-

+

-+

-+

-+

-+

ToT stages (time stamp and charge measure)

Cf

Cz

Rz

Rf

Cd-

+Current buer

Figure 411 Schematic of modified ToPix ASIC for strip readout

Figure 411 shows a possible adaption of the ToPixarchitecture that would take into account the afore-mentioned issues The input stage is a charge sen-sitive amplifier followed by a pole-zero cancellationnetwork This block can be seen as a current am-plifier with Gain CzCf By properly sizing Rf andCf one can limit the signal swing at the preampli-fier output so the saturation is avoided In thisway an adequate loop gain is always maintainedand the cross-talk is minimised The input stage isfollowed by a current buffer which provides addi-tional gain and drives the ToT block with properimpedance The ToT stage has the same topologyof the pixel cell in ToPix A constant current dis-charges the feed-back capacitance of an integratorThe start and stop time of the discharge are mea-sured by latching into local registers Gray-encodedwords provided by a counter which is common to allchannels Assuming the standard reference clock ofPANDA of 155 MHz a 10 bit resolution translatesinto a maximum dead-time of 64 micros The averagedead-time will be significantly smaller since higherinput charges which leads to longer signals are lessprobable By using more ToT cells per channel aderandomisation of the arrival time is achieved As-suming an average dead time of 64 micros (which is veryconservative for 10 bit resolution) and four digitis-ing cells one can accommodate a rate of 100 kHzper channel with an efficiency better than 995

Preliminary simulations of the architecture weredone using the parameters of the same CMOS pro-cess chosen for the prototypes of the pixel front-endelectronics The strip-optimised front-end shows anoise of 1000 electrons rms for a power consump-

tion of 1 mW per channel From the readout pointof view each channel which contains four indepen-dent digitising elements can be seen a ldquoshortrdquo pixelcolumn with four pixels Such similarities would ofcourse favour the development of a common readoutframework and the reuseability of different modulesand concepts between the two subsystems

STS-XYTER

The STS-XYTER is a development based on then-XYTER [5] It will make use of the same token-ring-architecture common to the n-XYTER It issupposed to be a dedicated front-end ASIC for thereadout of silicon strip detectors to be used in sev-eral FAIR experiments The front-end will feature aself-triggering architecture and is designed for lowpower consumption Therefore it will make use of aToT-based digitisation In order to handle the datarate in high occupancy environments a fast digitallink with a data rate up to 25 Gbps is envisagedTwo prototype ASICs - TOT01 [6] and TOT02 [7]- were produced and tested Further specificationsand parameters measured with first prototypes canbe extracted from table 44

FSSR2

The FSSR2 [8] is a 128-channel ASIC for siliconstrip readout It features a fast self-triggered read-out architecture with no analogue storage similarto the FPIX2 chip (the pixel front-end for the BTeVexperiment) where it was derived from Each ana-logue channel contains a programmable charge sen-


Parameter adapted ToPix STS-XYTER FSSR2

input pad pitch asymp 50 microm 50 microm 50 microm

channels per front-end 128 128 128

dynamical range 100 fC 15 fC 25 fC

ENC 1000 eminus 20 pF 700 eminus 28 pF 240 eminus+ 35 eminuspF

peaking time 6 ns 80 ns 65 ns

power consumption 08 mWch 12 mWch 40 mWch

trigger self-triggering self-triggering self-triggering

digitisation technique ToT ToT Flash-ADC

digitisation resolution 10 bit 4minus 6 bit 3 bit Flash-ADC

time resolution 185 ns 155 MHz 185 ns 155 MHz 132 ns time stamp

data interface e-link - SLVS up to 25 Gbps LVDS

number of data lines 1 pair 1 pair 1 to 6 pairs

slow control custom (serial) custom custom LVDS

process 013 microm CMOS UMC 018 microm 025 microm CMOS

radiation hardness gt 10 MRad 10 MRad up to 20 MRad

Table 44 Comparison of different front-ends for strip detectors The specifications for the STS-XYTER aretaken from prototype evaluations [6 7]

sitivity preamplifier a CR minus (RC)2 shaper withselectable shaping time a selectable base line re-storer and a discriminator The discriminator withan adjustable threshold simply yields a binary hitinformation Moreover each channel contains a 3-bit flash ADC with selectable thresholds to achievean amplitude information

43 Module Data ConcentratorASIC

431 Architecture

Each strip sensor module will have the front-endchips as well as a module controller on board Themodule controller serves as link between the front-end chips and the MVD detector data acquisition-system The module controller has to be able tohandle sensor modules of different sizes up to themaximum of 896 times 512 strips corresponding to thelarge double-sided sensors outlined in table 41 Themain tasks to be performed are

bull readout decoding and multiplexing of front-end chip data

bull buffering pedestal subtraction and optionalone-dimensional or two dimensional clustering

bull interfacing to DAQ via serial GBT [9] E-links

bull slow control functions (parameter transfermonitoring)

The module controller has to be finally realised asa fully digital ASIC in radiation-hard technologySince just one module controller is needed for eachsensor module the additional material budget thecurrent consumption and the thermal load of thiscomponent will be neglegible compared to that ofthe front-end chips The design of the module con-troller is currently in progress [10] To have a start-ing point it is assumed to use the n-XYTER plus aseparate ADC as front-end hardware In this case11 n-XYTERs and ADCrsquos are needed for the largeststrip sensor module The design is fully VHDL-based using Xilinx FPGA as technological platformThe modular structure of the design guarantees aminimum effort to incorporate different front-endsolutions or additional features in future versions

48frasl

frasl

frasl

48

48

Figure 412 Schematic of the Module Data Concen-trator


432 Implemented Feature ExtractionAlgorithms

The data frames from the different front-end chipscontain the time stamp strip-identifier and chargeinformation After subtracting the pedestal fromthe charge the frames are stored in a FIFO fromwhich they may be extracted in the raw format ora feature extraction may be performed by the mod-ule controller With the foreseen feature extractionstage enabled the module controller will continu-ously scan the FIFO output and collect frames of acertain time interval ∆t ∆t is determined by bit-truncation of a parameter-based number n of leastsignificant bits from the time stamp The corre-sponding frames are buffered and analysed for clus-ters with a minimal multiplicity m of next and nextbut one hits in the x- and y-strips of one sensormodule [11] If hits are detected the time stampthe centroid the width in the x- and y-plane thesum of charges and the multiplicity will be storedin the hit-register The contents of the hit registermay be transferred to the DAQ system or if en-abled passed to a 2-dimensional correlation stagethat generates a list of all possible x-y-combinationsof clusters For each possible combination a combi-natorial and a charge-correlation probability is cal-culated based on the utilisation of fast look up ta-bles particularly of error functions The resulting2d-clusters then are buffered in the output FIFOImportant clustering parameters like the minimalmultiplicity m the number n of truncated bits inthe time stamps the cluster thresholds or the cut-off threshold of the combinatorial probability areadjustable via the slow control interface

433 Implementation Status

The VHDL-design is expected to be finished in2011 First test were performed using device simu-lations The design is able to buffer up to 5 simulta-neous hits at the n-XYTER clock speed of 128 MHzBased on a Xilinx Spartan 6-device timing simula-tions resulted in approximately 150 ns for the multi-plexing of the 11 front-end devices (n-XYTER plusADC) The current consumption and thermal loadstrongly depend on the technology chosen for thefinal ASIC design and have to be evaluated laterSince the final decision concerning the front-endchip for the strip sensors has not been reached yetthe module controller design still has to be adaptedto the final front-end and transfered to the appro-priate ASIC technology basis

Figure 413 Basic concept for the hybridisation ofdouble-sided strip detectors

44 Hybridisation

441 Overview

The hybridisation of double-sided strip detectors isone of the technically most demanding tasks for thestrip part of the MVD Particular challenges arisefrom the electrical connection on both sensor sidesas described in chapter 533 Figure 413 gives aschematic illustration of the overall concept In con-trast to the hybrid pixel design readout electronicscan be placed outside the acceptance of the sen-sors Therefore appropriate adapter componentsare introduced to interconnect to sensor and thefront-end electronics Wire bonding is the defaultoption for all electrical connections from the sensorto the front-end chip Basically the layout of thereadout pads on the senor differs from the one onthe front-end This accounts for both the effectivecontact area as well as the pitch ie the interspac-ing of two neighbouring pads Technical limitationsof the wire-bonding process must be considered ifthe pitch size is smaller than 100 microm as it is thecase for the wedge sensors and the input channelsof the front-end These require parallel straight-lineconnections between corresponding pads which canbe achieved by the introduction of a fan-like struc-ture Such pitch adapter can be either integrated onthe hybrid carrier PCB or designed as an individ-ual component While the first option represents


Figure 414 Illustration of the hybridisation conceptfor the strip barrel part For the readout of the longstrips there are two major options The ganging of longstrips of adjacent sensors (a) or an individual readoutof all sensors (b)

an advanced technical solution the other one fol-lows the solution of other HEP experiments recentlyinstalled Finally the detector module must be in-tegrated onto a local support structure in order toachieve the required mechanical stability The localsupport structure includes a cooling pipe needed toprevent a heating of the readout electronics or thesensors

442 Basic Approach for the MVD

The different sensor geometry as well as the changedsensor orientation with respect to the beam pipelead to individual hybridisation concepts in the bar-rel and the forward part Figure 414 illustrates thebasic approach for the strip modules in the barrellayers A complete super-module contains severalsensors which are aligned in one row The car-bon fibre support structure is extended at the longedge thus leaving space for the positioning of thefront-end electronics Due to the stereo angle of90 the sensor is read out at two edges A flexiblecarrier structure is introduced to flip the front-endsat the short side by 90 to the support frame Theoverall number of readout channels can be furtherdecreased by connecting strips of neighbouring sen-sors However this option is technically more chal-lenging and may cause deteriorations of the over-all performance Therefore a fall-back solution isgiven by an individual readout of all sensors Aschematic cross section through a hybridised strip

module is shown in figure 415

The hybridisation concept of the strip disk part fol-lows the basic idea presented in [12] A disk super-module combines two successive sensors of the twolayers In contrast to the barrel sensors the readoutof both sensor sides is performed at the same sensoredge at the top Flexible adapter pieces facilitatea bending by 90 so that the front-end electronicscan be placed at the top plane in the gap betweenthe two sensors In this way a minimum radial oc-cupancy of the passive elements can be achievedWhile a flat support serves as contact surface forthe PCB carrier board a more sophisticated carbonsupport structure is needed to combine the readoutplane at the top and the two sensors to a compactand robust object An illustration of the hybridi-sation concept for the strip disks is given in fig-ure 416

443 Layout of Hybrid Carrier PCB

The material of the front-end electronics carrierstructure should satisfy the following requirements

1 low thickness at high structural integrity low-Zcompound

2 well defined dielectric constant up to frequen-cies of at least 1 GHz and low dielectric lossat the same time

3 radiation tolerance ie insignificant change ofproperties under irradiation in ionising as wellas non-ionising fields

From readily available materials only few fulfill allof these demands An optimal choice seems to beKapton1-Polyimide (chemical composition Poly-(44prime-oxydiphenylene-pyromellitimide)) since besidethe mentioned requirements additional advantageseg easy bondability with metal-films and wet-processing with standard-PCB-structuring tech-niques are applicable [13]

material kapton polyimide

dielectric thickness 254 microm

dielectric constant 34

dielectric loss tangent 0004

conductor type Cu

conductor thickness 18 microm

radiation length (min-max) 001minus 026 XX0

Table 45 Specifications of the hybrid carrier material

1 developed and patented by duPont


Figure 415 Schematic cross section of a strip hybrid structure in the barrel part The carbon frame carries thesensors (green) the flex-PCB with electrical structures and front-end electronics (red) The Kapton-Flex-PCB isglued with an overlap of ca 6 mm onto the sensor Connection to the latter is realised with aluminum wire-bondsThe Flex-PCB also serves as a fan-in adaptor to the pitch of the front-end ASIC which is mounted above thecooling pipe

Figure 416 Illustration of the hybridisation conceptfor the strip disks A flexible adapter (1) allows a bend-ing of the readout electronics at the top The super-module is supported by a sophisticated carbon struc-ture (2) and a flat structure at the top plane in betweenthe sensors (3) Cooling pipes (4) are placed below thefront-end

For the hybrid carrier PCB a two layer layout isforeseen with the parameters given in table 45 Thevalue for radiation length for this material is in gen-eral very low but may peak up to ca 025 XX0

locally when copper traces or larger areas on bothsides overlap

The fan-in routing of the default 130 microm pitch struc-ture to the front-end pitch can be processed by stan-dard PCB manufacturing techniques Figure 417

Figure 417 Layout detail of pitch adaptor hybrid

shows a layout of a prototype for the connection toalready existing n-XYTER front-end ASICs whichis being produced and evaluated at the momentThe bonding pads at the top edge are placed in twostaggered rows at the sensor pitch The next step isthe interconnection of this hybrid prototype to thesensor by directly glueing the PCB with an overlapof 67 mm onto the sensor such that the innermostAC-pads of the sensor and the bonding pads on theflex-PCB face each other and can be connected viabonding wires


444 Interconnections

Basically there are three different subjects tobe considered as interconnection between differentcomponents of a super-module

bull Electrical connectionsElectrical connections are needed to supply thesensor and the electronics to lead the chargesignal of individual strips to an input channelof the front-end electronics as well as to trans-mit data signals on the super-module There-fore wire bonds with a thickness of a fewtens of micrometres are foreseen With respectto a minimised effective radiation length alu-minium bonds will be preferably used How-ever there are technical challenges in particu-lar for very thin wires which are taken for thecharge signals Therefore gold wires are thedefined standard for some of the wire-bondingtechnologies

bull Mechanical connectionsAll the components of a super-module will beglued together The choice of the glue dependson kind of interconnection and thus must bein accordance with given specifications on theelectrical and thermal conductivity the expan-sion coefficient and the effective material bud-get In any case it must withstand the ex-pected radiation dose For the interconnectionof the front-end to the carrier boards a highthermal conductivity of the glue is needed Onthe other hand a too high material load mustbe prevented

bull External connectionsConnectors are needed for the overall integra-tion of a super-module Cooling connectorsmust allow for the change from a stable pipe toa flexible tube Moreover separate connectorsfor the power supply and the data transmissionhave to be inserted Finally the mechanicalconnection to external holding structure mustbe ensured with a high reproducibility

Most of the parts mentioned above related to in-tegration aspects are very similar for the pixel andthe strip part Therefore more details on specificmaterials eg glues can be found in chapter 5

445 Test Assembly

First evaluation studies of the hybridisation conceptfor the strip part have been performed with a lab-oratory test setup [14] for the characterisation of

double-sided silicon strip detectors and connectedreadout electronics Therefore square sensors witha side length of 2 cm and a thickness of 300 microm [15]were used in combination with the APV25-S1 read-out chip [16] Specifications of both componentsdiffer partly from PANDA requirements Howevershape and dimensions of the implemented test sen-sor are similar to the one defined for the MVD bar-rel sensors Moreover the stereo angle of 90 be-tween the strips of both sensor sides is compliantThe strip pitch of 50 microm falls short of the small-est value of 675 microm defined for the wedge sensorsTherefore the hybridisation of appropriate detectormodules facilitates an evaluation of technical detailssuch as mounting procedures and the electrical con-nection via wire bonds at very small dimensions

Figure 418 Photograph of a finally assembled detec-tor module for the laboratory test setup Main compo-nents are (1) Sensor (2) Pitch-adapter (3) Front-endchips (4) High-density connector for power supply slowcontrol data output (5) HV connection for sensor sup-ply

A photograph of the completed detector module isshown in figure 418 The sensor is glued to L-shaped circuit boards thus allowing a double-sidedreadout with one common design A ceramic pitch

Figure 419 Photograph of the fabricated pitchadapter and schematics of the pad geometry


Figure 420 Results of long-term measurements with a 90Sr source on the laboratory setup Correlation ofdeposited energy (left) for both sensor sides (p-side and n+-side) and 2D imaging of a SMD device placed on topof the sensor and acting as an absorber for the β-electrons (right)

adapter has been fabricated which leads from thereadout pitch of the sensor to the needed struc-ture of the APV25 front-end chip with a pitch ofjust 44 microm One photograph of the fabricated pitchadapter can be found in figure 419 along with theschematics of the pad geometry

Extended studies with a series of detector moduleshave been performed proving the full functionalityof the setup Moreover there are no indications ofan impact of the assembly procedure on the finaldetector performance Selected results of the mea-surements are collected in figure 420

References

[1] CiS Forschungsinstitut fur Mikrosensorik undPhotovoltaik GmbH httpwwwcismstdeOct 2011

[2] Rose Collaboration (RD48) Third RD48 Sta-tus Report Dec 1999

[3] M Moll E Fretwurst and G Lindstrom Leak-age current of hadron irradiated silicon detec-tors material dependence Nucl Instr MethA426(1)87ndash93 1999

[4] J Ellison and DOslash collaboration Silicon mi-crostrip wedge detectors for the DOslash silicontracker Nucl Instr Meth A34233ndash38 1994

[5] AS Brogna et al N-XYTER a CMOS read-out ASIC for high resolution time and ampli-tude measurements on high rate multi-channelcounting mode neutron detectors Nucl InstrMeth A568 Issue 1pp 301ndash308 2006

[6] K Kasinski R Szczygiel and P GrybosTOT01 a Time-over-Threshold based readoutchip in 180 nm CMOS technology for siliconstrip detectors JINST 6 C01026 2011

[7] K Kasinski R Szczygiel and P Grybos Evo-lution of a prototype silicon strip detector read-out ASIC for the STS GSI Scientific Report2010 page 232 May 2011

[8] R Yarema et al Fermilab Silicon Strip Read-out Chip for BTeV IEEE Trans Nucl Sci52 No 3pp 799ndash804 2005

[9] P Moreira et al The GBT Project In Proceed-ings of the Topical Workshop on electronics forparticle physics pages 342ndash346 Paris FranceSept 21-25 2009

[10] H Sohlbach Readout controller design for theMVD Si-strip sensors March 2011 XXXVIPANDA Collaboration Meeting

[11] Karsten Koop FPGA-basierte Auslese vonSilizium-Streifen-Detektoren Diploma ThesisUniversitat Bonn 2009

[12] JJ Steijger The lambda wheels a silicon ver-tex detector for HERMES Nucl Instr MethA453 (12)98ndash102 2000


[13] duPont General Specifications of KaptonPolyimide Films Jun 2010 httpwww2

dupontcomKaptonen_US last visited onJul 3 2011

[14] Th Wurschig L Ackermann F Kruger RSchnell and H-G Zaunick Setup of a test-station for double-sided silicon microstrip de-tectors PANDA-MVD-note 005 2010

[15] D Betta et al Development of a fabricationtechnology for double-sided AC-coupled sili-con microstrip detectors Nucl Instr Meth460306ndash315 2001

[16] M Raymond et al The CMS Tracker APV25025 microm CMOS Readout Chip Paper presentedat the 6th workshop on electronics for LHC ex-periments Krakau Poland Sept 2000

81

5 Infrastructure

51 Optical Data Transmission

The maximum expected data rate for the pixelMVD is estimated to be 450 Mbitchip The8b10b encoding on the output data could increasethe number to sim 600 MbitchipDue to the absence of a trigger signal no triggermatching logic can be used to reduce the amount ofdata already at the front-end chip level Moreoverthe track distribution over the barrel and the diskscan vary significantly depending on the type of thetarget thus not allowing an optimisation based onthe detector position These constraints make thepixel MVD data transmission extremely demandingin terms of required bandwidthAnother stringent constraint is due to the reduceddimensions of the MVD and its closeness to the in-teraction point which limits the number of cablesthat can be accommodated and tolerated (for ma-terial budget issues)These constraints lead to the choice of using highspeed serial links and to convert the data transmis-sion from electrical to optical as close as possible tothe detector in order to profit from the high datarates allowed by the optical fibersUnfortunately the use of off-the-shelves componentsfor this application is not straightforward becauseof the radiation environment where the componentswill be placed However the problem is commonto most of the HEP community and in particu-lar for the upgrades of the LHC experiments Inthis context two research projects named GigaBitTransceiver (GBT) [1] and Versatile Link (VL) [2]are ongoing in order to develop a radiation toleranttransceiver (the GBT) and to qualify optical com-ponents for the radiation environment (the VL)Theconnection between the MVD and the DAQ is goingto make profit of this development and therefore todesign a front-end electronic compatible with theGBT interface Figure 51 shows the scheme pro-posed by the GBT and VL projects

511 GigaBit Transceiver

The goal of the GBT project [1] is to produce theelectrical components of a radiation hard opticallink as shown in figure 51 One half of the systemresides on the detector and hence in a radiationenvironment therefore requiring custom electron-ics The other half of the system is free from ra-

diation and can use commercially-available compo-nents Optical data transmission is via a system ofopto-electronics components produced by the Ver-satile Link project The architecture incorporatestiming and trigger signals detector data and slowcontrols all into one physical link hence providingan economic solution for all data transmission in aparticle physics experimentThe on-detector part of the system consists of thefollowing components

bull GBTX [3] [4] a serialiser-de-serialiser chipreceiving and transmitting serial data at48 Gbs It encodes and decodes the datainto the GBT protocol and provides the inter-face to the detector front-end electronics

bull GBTIA [5] a trans-impedance amplifier re-ceiving the 48 Gbs serial input data from aphotodiode This device was specially designedto cope with the performance degradation ofPIN-diodes under radiation In particular theGBTIA can handle very large photodiode leak-age currents (a condition that is typical forPIN-diodes subjected to high radiation doses)with only a moderate degradation of the sen-sitivity The device integrates in the same diethe transimpedance pre-amplifier limiting am-plifier and 50 Ω line driver

bull GBLD [6] a laser-driver ASIC to modulate48 Gbs serial data on a laser At presentit is not yet clear which type of laser diodesedge-emitters or VCSELs will offer the besttolerance to radiation [2] The GBLD was thusconceived to drive both types of lasers Thesedevices have very different characteristics withthe former type requiring high modulation andbias currents while the latter need low bias andmodulation currents The GBLD is thus a pro-grammable device that can handle both typesof lasers Additionally the GBLD implementsprogrammable pre- and de-emphasis equalisa-tion a feature that allows its optimisation fordifferent laser responses

bull GBT-SCA [7] a chip to provide the slow-controls interface to the front-end electronicsThis device is optional in the GBT systemIts main functions are to adapt the GBT tothe most commonly used control buses used


Figure 51 GBT and VL scheme LD Laser Driver TIA TransImpedance Amplifier PD Photo Diode TRxTRansceiver

HEP as well as the monitoring of detector envi-ronmental quantities such as temperatures andvoltages

The off-detector part of the GBT system consistsof a Field-Programmable-Gate-Array (FPGA) pro-grammed to be compatible with the GBT protocoland to provide the interface to off-detector systemsTo implement reliable links the on-detector compo-nents have to be tolerant to total radiation dosesand to single event effects (SEE) for example tran-sient pulses in the photodiodes and bit flips inthe digital logic The chips will therefore be im-plemented in commercial 130 nm CMOS to ben-efit from its inherent resistance to ionizing radia-tion Tolerance to SEE is achieved by triple mod-ular redundancy (TMR) and other architecturalchoices One such measure is forward error correc-tion (FEC) where the data is transmitted togetherwith a Reed-Solomon code which allows both errordetection and correction in the receiver The formatof the GBT data packet is shown in figure 52 Afixed header (H) is followed by 4 bits of slow controldata (SC) 80 bits of user data (D) and the Reed-Solomon FEC code of 32 bits The coding efficiencyis therefore 88120 = 73 and the available userbandwidth is 32 Gbs FPGA designs have beensuccessfully implemented in both Altera and Xilinxdevices and reference firmware is available to usersDetails on the FPGA design can be found in [8]

52 Off-Detector Electronics

The PANDA DAQ System architecture defines 3multiplexing and aggregation layers before the de-tector data are sent to the Level 1 Trigger farm im-

Figure 52 GBT data packet format

plemented by the Compute Nodes in ATCA cratesThe first layer does the multiplexing of severalfront-end input links into one outgoing link In ad-dition to this information on beam structure (2 microsbursts and 500 micros super bursts) and global clock isreceived by the SODA system and propagated tothe front-end modules in order to provide propertime-stamping of the detector data In two fur-ther aggregation stages data are grouped into datablocks corresponding to one burst and super blockscorresponding to one super burst

Due to its extreme data rates the MVD provideshigh challenges for this architecture Assuming thatthe MVD module controllers and service boards(with the GBT) are combined such that the GBTlinks are almost fully loaded the MVD MultiplexerBoards (MMBs)will do a 3-to-1 multiplexing with 3optical links from the GBTs on the service boardsand one optical uplink with 10 Gbs Since themaximum user data rate for the GBT is 328 Gbsall three links from the service boards can be fullyloaded simultaneously Due to the high output ratethe MMBs will be designed as MicroTCA-Boards


Figure 53 Architecture of the lower levels of thePANDA DAQ system

This allows direct insertion into the Compute Nodeversion 3 and skipping of the SuperBurst build-ing stages Implementation of test systems is sup-ported as well

The MMB will not use the GBT but will imple-ment the GBT protocol with a commercial FPGAThe Xilinx Virtex 6 is foreseen for this purpose butthe final decision will be done during the hardwaredesign phase Figure 54 shows the main compo-nents of the MMB SODA provides global clock andtiming as well as information on bursts and superbursts which will be propagated by the FPGA viathe optical links to the service modules using thetriggeramptiming interface component of the GBTThis information in combination with the referenceclocks and reset pulses generated by the GBT al-lows proper time stamping of the FE ASIC dataVia the FPGA the GBT-SCA is accessible in orderto do environmental monitoring Setting the pa-rameters of the FE ASIC (eg thresholds) is donevia the DAQ interface of the GBT and not with theGBT-SCA On the MicroTCA interface the PCIeprotocol (with 4 lanes) is used in the fat pipe ThePCIe protocol is implemented by the FPGA andgives access to the above-mentioned managementand control functions Also the detector data canbe sent via the MicroTCA interface instead of theoptical uplink In this case the average load overall 3 links from the GBTs must not overcome 80The module management controller (MMC) is im-plemented by a Microcontroller which communi-cates with the shelf manager via I2C

In normal operation it is foreseen that the MMBswill reside in MicroTCA crates Each MicroTCAcrate will contain one commercial off-the-shelfCPU-board for management and control purposesLinux will run on the CPU-board enabling comfort-

Figure 54 Main components of the MMB (showingnetto data rates of the link without GBT protocol or8b10b overhead)

able local and remote access to the MMBs via thePCIe link on the MicroTCA backplane The shelf-monitoring functions of MicroTCA support addi-tional remote monitoring and control operations

53 Power Supply System

531 Introduction

The power supply system for the whole MVD has toprovide low voltage for a total number of 2130 read-out circuits including pixel and strip electronicsand 346 controller or transceiver circuits to trans-mit the data towards the DAQ

Besides the power supply system has to providehigh voltage for a total number of 136 super mod-ules including both pixel and strip sensors

532 Powering Concept for the PixelPart

For the hybrid pixel part the smallest unit con-sidered for the segmentation of the power supplysystem is a super module that can include upto 6 readout chips (ToPix) and a transceiver chip(GBT) For each basic unit the design of the powersupply system has to provide two independent float-ing channels for the analogue and digital sectionswith sensing feedback for the voltage compensationdue to the cable voltage drop

With respect to the power supply system the re-quirements for the ToPix readout chip are a chan-nel for the analog circuits with a nominal voltageVa = 12 V an upper voltage limit Val = 14 Va nominal current Ia = 200 mA an upper currentlimit Ial = 220 mA and a channel for the digitalcircuits with a nominal voltage Vd = 12 V an up-per voltage limit Vdl = 14 V a nominal current


Id = 12 A an upper current limit Idl = 14 AThese numbers are evaluated after the measurementon the ASIC prototype ToPix v3 and extrapolatingthe result to a full size chip with the complete arrayof cells that affects mainly the analogue currentand the entire set of columns that affects the digitalcurrent A significant amount of the digital currentis due to the differential drivers (7 outputs) and re-ceivers (9 inputs) constituting the interface for thechip that at present is considered completed anddoes not require to be increased as in the case of thecurrent for the end of column blocks Currently theexpected currents are evaluated considering a read-out circuit running with a system clock 160 MHzfast and a pre emphasis capability to drive the longtransmission lines while working at 80 MHz or re-leasing the boosting could decrease the total currentconsumption of the 20 or 5 respectively There-fore sensible current limits for the power supplyhave to be selected to protect the readout with-out requiring too much power that could lead to anoversised system for this purpose since the presentfigures constitute the worst case an increment ofjust 10 for all the parameters has been chosen

Power Regulators

As power regulator the linear regulator solution hasbeen excluded due to the low efficiency in the powerconversion while a switching DC-DC solution hasbeen chosen because it has an higher efficiency andcan be based on the project currently under de-velopment at CERN for the upgrade of the LHCexperiments

The switching converter needs to store the energyeither by the magnetic field (using inductors) or bythe electrical field (using capacitors) at present themain project is addressing the architecture with in-ductors that allows larger efficiency although thearchitecture with capacitor looks very attractivedue to the small size of the whole circuit Thefull DC-DC converter is a kind of hybrid basedon a small PCB containing few passive compo-nents mainly the inductor and an ASIC providingthe switching transistors and the control logic tomonitor the correct behaviour [9] Currently thereare different prototypes featuring integrated circuitsmade with different technology to test their hard-ness with respect to the radiation dose but in anycase all of them foresee an input voltage aroundVi = 10 V an output voltage Vo = 12 V compat-ible with the new readout circuits and an outputcurrent of roughly Io = 3 A

This type of inductor based converter also called

ldquobuck converterrdquo requires particular care whenworking in a high magnetic field and for this rea-son it cannot make use of ferrite core coils since itrisks the saturation but an air core coil is preferredalthough an increment in the size is expected An-other issue related to the inductor is the radiatedswitching noise and his electromagnetic compati-bility with the system for this purpose the coil isshielded by a small box with the conductive surfaceand the geometrical dimensions become in the orderof 30times 10times 10 mm3

At present the CERN development is selecting thebest technology for the ASIC production from thepoint of view of the radiation hardness but abackup solution has been found in the 350 nmOn Semiconductor that has been already qualifiedfor this application The other technology optioncomes from the 250 nm IHP but since it is a moreadvanced solution it has not yet reached the neces-sary level of radiation hardness and it isnrsquot enoughtolerant against the single event burnout or the dis-placement damage Specifically these DC-DC con-verters have been tested to cope with the environ-ment of an experiment for high energy physics con-sidering a total integrated dose of around 1 MGy aparticle fluence of roughly 1015 n1MeV eq cmminus2 anda magnetic field of about 4 T

533 Powering Concept for the StripPart

For the strip part the foreseen powering scheme forone super module contains separate low voltage sup-ply paths for the n-side and the p-side readout elec-tronics respectively since the n-side front-end elec-tronics is operated without galvanic reference to acommon ground in a floating regime This avoidshigh parasitic currents in the case of breakdown ofthe thin insulation barrier between n-side implantsand metallisation (pinholes) of the sensors whichwould otherwise result in the loss of the completesensor Thus a total of 4 independent low volt-ages (one analogue and one digital for either sensorreadout side) and 4 high voltages for each sensorbias have to be supplied The low voltage paths areconnected with sense feedback to compensate forthe voltage drop on the cables Assuming a maxi-mum of 1 W power dissipation per front-end at anoperation voltage of 33 V a value of 53 A can beestimated for the maximum supply current on thelow voltage lines The current drawn on the biasvoltage lines is negligible even with the maximumapplied irradiation dose at the end of the PANDAlifetime and amounts to at most 2 mA per sensor


Sensing lines are thus not required for this supplybut a good filtering with very low cutoff frequenciesis mandatory on the module level

54 Cables

A special feature of the PANDA experiment comesfrom the triggerless DAQ that requires a very highoutput bandwidth because to the continuous flux ofdata that cannot exploit the de-randomisation com-ing from an external trigger at a lower rate Besidessince at present the only way out from the MVD isin the backward region the cabling becomes a seri-ous issue because of the amount of material neededto take out all the data from the sensitive volume

541 Requirements

Due to the average annihilation rate of 20 MHz aflux of 3 middot 106 particles per second and chip is ex-pected in the hottest region the forward disks Thecurrent design for the pixels considers a ToPix read-out circuit managing a 100 microm thick sensor a dataword of 40 bit per event a variable luminosity struc-ture due to the peak in the beam profile and thecrossing of the pellet target therefore the readoutchips in the middle of the disks have to deal with adata rate of 450 Mbits each

The present solution for the architecture of theserial link comes from the Giga Bit Transceiver(GBT) project under development at CERN for theupgrade of the experiments that will be running atthe Super LHC (SLHC) The GBT is foreseen to beinterconnected to the readout circuits by an Elinkthat is an electrical interface implementing a serialconnection

542 Signal Cable

To deal with the low material budget for the vertexregion the use of copper cables is excluded and onehas to employ aluminium cables In this case con-sidering that aluminium has a lower atomic numberthan copper even if it has a higher resistivity anoverall gain is obtained in the material budget be-cause the aluminium radiation length is 889 mminstead of copper radiation length that is 144 mm

From the physical point of view each logical sig-nal going to or coming from ToPix chip is im-plemented by a differential pair to be less sensitiveagainst the noise that can be induced at the sametime on each single line and that will be canceled

after the subtraction between the two phases Inparticular these connections are implemented as atransmission lines using the microstrip techniqueand are composed as a sandwich of two aluminiumconductive layers separated by a polyimide insula-tor layer [10] On one side the two tracks compos-ing the differential pair are laid down while on theother side the metal plane constitutes the referenceground

Al Prototypes

Several kinds of cable prototypes were made andtested differing either in the manufacturing tech-nology or in the physical layout For the first choicethe standard solution is a laminated sheet over anpolyimide layer that is worked in a way similar to aPCB etching the metal from the area where it mustbe removed

The other manufacturing solution is an aluminiumdeposition on top of an insulator film that is ob-tained dispensing the metal just in the requiredarea defined by the layout pattern by the Physi-cal Vapour Deposition (PVD)

Folded Layout

The first samples were produced to test both thetechnologies by using the metal lamination fromthe Technology Department at CERN and succes-sively the aluminium deposition from the Techfabcompany near Torino The CERN prototype has astack of up to two aluminium foils with a thick-ness of 15 microm on both sides of the plastic film51 microm thick produced by DuPont under the com-mercial name of Pyralux Various layout prototypeswere produced with different widths and spacing of100 microm + 100 microm 150 microm + 150 microm and 200 microm+ 200 microm respectively

The Techfab prototype features a cross sectionof two aluminium depositions with a thickness ofroughly 7 microm around the 50 microm thick plastic sheetproduced by DuPont as common kapton In thiscase a sample with a unique layout showing 150 micromwidth and 150 microm spacing has been obtained Dueto some issues with the wire bonding procedure atry has been performed with an aluminiumsiliconalloy to improve the bond strength The sampleswith the folded layout have been used to performan irradiation test by neutrons at the LENA nu-clear reactor in Pavia with a total fluence greaterthan what is expected for the PANDA lifetime andno significant variations have been observed


Straight Layout

The low mass cables have been designed implement-ing the connections as a transmission line using themicrostrip technique on DuPont materials and man-ufacturing them in a way similar to a PCB Cur-rently there are aluminium microstrips producedby the Technology Department at CERN with alayout including 18 differential pairs for a totalnumber of 36 tracks at present there are two ver-sions differing for the nominal width that is 100 micromor 150 microm and the spacing that is 100 microm or 150 micromrespectively All of them have the insulator coveron the bottom side to protect the ground planewhile one more parameter introducing another dif-ference on the available samples is the presence orthe absence of the insulator cover laid on the topside where the differential strips are implemented(figure 55) Summarising at present there are 4different kinds of aluminium microstrips 1 m longunder study differing for the pitch size or for thetop covering

Figure 55 Low mass cables implementing aluminiummicrostrips

Results

From a static point of view the aluminium mi-crostrips were tested to measure the resistance ofeach track between its ends and the capacitancereferenced to the ground plane moreover the uni-formity for these parameters was evaluated (fig-ure 56) The measurement has been performedbonding both cable ends to the test station to re-duce the contact problems and no dependence be-tween static parameters and track position has beenobserved Instead as expected the results showthat there is a clear relation between the linear re-sistance and the track width that is around 12 Ωfor the large strips (150 microm) and about 17 Ω for

Figure 56 Linear resistance measured for the first 20tracks of each sample

the narrow ones (100 microm) Regarding the dynamicbehavior a test bench has been setup with someprototype circuits implementing the Scalable LowVoltage Signaling (SLVS) standard to measure thetotal jitter and its main components like the de-terministic jitter or to perform a Bit Error RatioTest (BERT) as a function of the data rate [11]Other tests were performed to evaluate the crosstalk effect with the transceiver prototypes and tomeasure the total jitter produced by just the alu-minium cables without introducing the componentdue to the SLVS circuits The present solutionconsiders that signals running on the low mass ca-bles comply with the SLVS differential standardwhich foresees for each line a Vlow = 100 mV anda Vhigh = 300 mV against the LVDS differentialstandard which typically foresees for each line aVlow = 105 V and a Vhigh = 145 V The setupfor the measurement is composed of the aluminiumcable under test that is glued by the shield on twoboards specifically designed while the interconnec-tions are made with a standard wire bonding Thetest was performed by a pulse generator configuredto produce a Pseudo Random Bit Sequence (PRBS)with a pattern length of p = 223 minus 1 bits and out-put levels according the SLVS standard In thisway the performance of the aluminium microstripcan be evaluated studying the analog shape of thesignals coming out of cable without the insertionof any distortion attributable to the transceiversThe 4 samples measured so far are called w1000cw1002u w1500u and w1501c where the firstnumber represents the track width in micrometerand the last letter represent the condition to be cov-ered or uncovered with the insulator film The totaljitter has been measured and evaluated in Unit In-terval (UI) that represents the ideal period of a bitirrespective of the link speed allowing the relative


comparison between different data rate (figure 57)Since the commercial serialiser and deserialiser areconsidered safe when their total jitter came up to athreshold of Tj = 03 UI it is possible to concludethat the aluminium microstrips with track widthw = 100 microm are able to run up to a data ratef = 600 Mbits while the differential cables withtrack width w = 150 microm can work up to a data ratef = 800 Mbits

Figure 57 Total jitter evaluated for just the mi-crostrips without any transceivers

An useful tool to evaluate the total jitter with re-spect to the unit interval is the eye diagram seefigure 58 that is a composite view of the signalmade by the overlapping of many acquired wave-forms upon each other corresponding to bit periodsrepresenting any possible logic states and transi-tions

Figure 58 Example of an eye diagram

Moreover analysing the eye diagram height at theend of the flexcable relative to the input PRBS the

differential amplitude can be evaluated and it re-sults in final value of roughly 300 mV instead of thenominal value of about 400 mV with an attenua-tion that is due to the tracks linear resistance Thetest setup has allowed an evaluation of the cross talkkeeping a victim differential pair biased in a steadystate and driving the aggressor differential lines onthe left and right sides of the first one by the samesignal but a very light effect has been reported andprobably this is due to the little swing of the SLVSstandard and the relatively low transition time ofthe signals connected to the high capacitive load

543 Power Cable

Each ToPix circuit is expected to consume around14 A and since it requires a supply voltage at 12 Vthat means a total power of 168 W typically corre-sponding to a relative power of 979 mWcm2 How-ever two independent lines have to be distributedfrom the power converter to separate the analogpart from the digital part to reduce the risk of mu-tual interference

Also in this case the material budget requirementrepresents a constraint that prevents the use ofcopper conductors inside the active volume of theMVD then a solution with aluminium was investi-gated At present a sample of enameled wire madeof copper clad aluminium has been evaluated to ver-ify that it is suitable for the routing in a very limitedspace reducing the voltage drop to roughly 100 mV

55 Mechanical Structures

The whole MVD is composed by four mechanicallyindependent sub-structures

bull pixel barrels

bull pixel disks

bull strip barrels

bull strip disks

Each of them is assembled independently thengathered on a stand and positioned on a holdingframe Figure 59 shows a longitudinal cross sectionof the MVD together with the cross-pipe sector

551 Global Support The Frame

The frame is the main structural element of thewhole MVD Its purpose is to suspend and keep in


Figure 59 Longitudinal cross-section of the MVD and the cross-pipe sector

position all the sub-structures with a relative po-sitioning error of less than 100 microm as well as toconnect the MVD to the external world (ie thecentral tracker) The frame also acts as supportfor the services from the inner part of the MVDThe frame has a sandwich structure made by ahigh modulus composite of unidirectional carbonfiber cyanate ester resin and foam as core materialThe cyanate ester resin systems feature good behav-ior and high dimensional stability when immersedinto a temperature-humidity varying field The useof a conventional epoxy system would have draw-backs due to excessive moisture absorption andresidual stress As compared to epoxies cyanateester resins are inherently tougher and they havesignificantly better electrical properties and lowermoisture uptake [12] [13] The main propertiesof the M55JLTM110-5 composite material are re-ported in table 51

A finite elements analysis has been performed inorder to define the minimum number of plies nec-essary to achieve the requested stiffness Accord-ing to the simulation the frame has been designed

Tensile Modulus E11 (GPa) 310

Tensile Strength σ11 (MPa) 2000

Poissonrsquos ratio ν12 310

CTE fibre (ppmC) minus11

CTE resin (ppmC) 60

Moisture uptake (155 d - 22C - 75 RH) 14

Table 51 M55JLTM110-5 Composite properties

including two skins each composed of four plieswith a quasi-symmetric stacking sequence of uni-directional M55J carbon fiber prepreg and 3 mmof Rohacellreg51IG as core material A set of eightsemi-circular 1 mm thick reinforcing ribs is embed-ded into the structure These ribs will be realisedwith the same composite material used for the ex-ternal skins The final thickness of the frame is4 mm

Figure 510 Maximum vertical displacement of theframe under static load A safety factor of 2 is applied

Figure 510 shows the behavior of the frame un-der gravitational load Extra loads are applied dis-


tributed on the end-rings to take into account asafety factor of 2 The simulation shows a maxi-mum displacement in vertical direction of less than100 microm

The frame holds the four sub-structures contain-ing the silicon devices in their place Two ringsglued at the ends of the frame act as connectors be-tween the sub-structures and the frame itself Therings are made of mattglassepoxy laminate com-posite DurostoneregEPM203 pre-machined and fin-ished after gluing in order to achieve the final di-mensions and the requested accuracy A set ofproper tools is necessary for these operations

The relative position of the two halves is ensured bymeans of a three cam system Figures 511 512 and513 show the ancillary components of the systemTwo cylindrical cams are located on the upper endsof one half frame whereas a third one is located atthe lower end A sphere is positioned on top of thecams On the other half frame there are insertswhich are embedded into the rings One insert onthe top has a ldquoVrdquo shape while a second one alsoon the top but at the opposite end is ldquoUrdquo-shaped

Figure 511 Cams system Cylindrical cam on upperend for fixing the half frame

Figure 512 Cams system The V insert

Figure 513 Cams system The U insert

The upper cams pass through the inserts and fix theposition along the Z axis The third cam locatedon the bottom of the half frame fixes the radialposition All the ancillaries are made by hardenedaluminium alloy (Hotokol) A similar system allowsthe MVD to be suspended from the Central Frameand to be fixed with three springs The springs areshaped so to get a two-components locking forceone component (X) is on the horizontal plane fixingthe radial position while the second is on the ver-tical plane downward oriented and defines the Yand Z components Finally the position is achievedby means of the V and U shaped inserts embeddedinto the Central Frame structure

This system allows the control of the position to beachieved so that re-positioning is possible with ahigh degree of accuracy that we can evaluate to beless than 10 microm [14]

552 The Pixel Support Structure

Pixel Barrels

The pixel barrels are a collection of two cylindri-cal layers of detectors placed around the interactionpoint and coaxial with the beam pipe axis

The first innermost layer is composed by 14 super-modules while the second layer by 28 Each super-module is a linear structure of pixel sensors hostingfrom two to twelve elements glued on a mechani-cal support made of a carbon foam (POCO HTCand POCO FOAM) thin layer and a Ω-shaped sub-strate realised with a M55J carbon fibre cyanateester resin composite The cooling tube is embed-ded between the carbon foam and the Ω substrate


Figure 514 Structure of a pixel barrel super module

Figure 514 sketches the structure of a super moduleand figure 515 shows its transverse section

A plastic block glued on one end of the Ω substrateacts as reference The block has two precision holesthrough which two precision screws fasten the supermodule to the support structure

Figure 515 Transverse section of a pixel barrel supermodule

The support structure is an assembled set of sub-structures Due to the strong geometrical con-straints the shape of the support structure isformed by the surfaces of two semi-truncated conesinterconnected by two reference rings The ref-erence rings contain the positioning holes for thesuper-modules and allow the precise positioning ofthe detectors

The support structure shown in figure 516 is sus-pended from the frame by a reference cross flangeIn figure 517 the half barrel of the second layerof pixel is shown assembled on the cone support

Figure 516 Support structure of the pixel barrel

In each barrel the staves are arranged at two dif-ferent radius and the overlap of the active areas issim 03 mm

Pixel Disks

The disks are six planar structures arranged at dif-ferent distances from the interaction point Eachdisk is divided vertically in two parts called halfdisks In this way it is possible to assemble thewhole set around the pipe Each disk is composedby a planar support structure by detectors mod-


Figure 517 Half barrel assembled on the cone support

ules and by cooling tubes The carbon foam planarstructures also act as cooling bridges The coolingtubes are embedded into the disks while the detec-tors modules are glued on both sides of each disk inorder to get the largest possible coverage

The carbon foams both the POCO HTCand the POCO FOAM are supplied inrectangular plates Typical dimensions are300 mmtimes 300 mmtimes 13 mm With the goal ofcutting the shapes for both the barrel elements andthe disks elements in the most efficient way withminor possible machine rejection in mind differentproduction methods have been investigated Themost efficient way is to cut the shapes from thecarbon foam plates with wire EDM

Figure 518 The cut of the shape of the disk from theplate with wire EDM

As general philosophy the shapes are cut from theplate (figure 518) and subsequently sliced-up (fig-ure 519)

Figure 519 Machining of the disks elements the halfdisk shapes are cut with wire EDM and then sliced-up

The wire EDM operates with the material to be cutimmersed into water The foam acts as a ldquospongerdquoAfter the wire cut a big quantity of water was em-bedded into the foam element In order to removethe water absorbed the element was baked in anoven for 48 hours at 60 C Before the baking thefoam element weight was 1512 g while after 48hours of baking the weight became 1312 g as ex-pected for the dry element

The set of six half disks is assembled to obtain a halfforward pixel detector (see figure 520) The rela-tive distances between disks are ensured by spacersglued on the disk surfaces The final position of eachhalf disk is surveyed and referred to a set of exter-


nal benchmarks fixed on a suspender system whichfastens the half forward detectors to the frame aswell

Figure 520 Sketch of the set of half disks composinghalf of the pixel forward detector

553 Support Structures of the StripPart

Strip Barrels

The main holding structure of the strip barrelpart consists of two half-cylinders with a radius of114 cm and an overall length of 493 cm Theyare made out of a sandwich structure of two layersof carbon fibers (M55J) with a thickness of 02 mmeach and a 2 mm core of Rohacell foam These half-cylinders are attached to the global support struc-ture of the MVD via 4 connectors at the upstreamend of the detector (figure 522) Sawtooth shapedstructures on the inside and the outside form a sup-port surface for the sensor elements sitting on socalled super-modules A super module consists offour 6 cm long sensor modules and one or two shortmodules with a length of 35 cm In addition asuper module houses the readout electronics inter-connections and the cooling system The supportstructure is made out of the same sandwich struc-ture used already for the support cylinders Under-

neath the readout electronics the Rohacell core isreplaced by a carbon foam (POCO HTC) with highthermal conductivity and an embedded cooling pipeof 2 mm diameter The edges of the structure arebent to increase the stiffness of the structure (fig-ure 523) The fixation of a super module on thesawtooth rings is done via special bearing pointswhich allow a precise positioning and a compen-sation for different thermal expansion coefficientsThe overall concept is based on the ALICE ladderrepositioning system with a precision better than6 microm [14]

Strip Disks

The two forward disks are made of two double layerhalf-disks which are separated along the verticalaxis They have an outer radius of 137 mm an innerradius of 75 mm and a spacing of 70 mm betweenthem The basic building block of the double-disk isa module which consists of two wedge shaped sen-sors which share a common PCB for the readoutelectronics and the cooling pipes They are con-nected to a carbon foam sandwich structure whichholds the sensors the PCB the two cooling pipesand does the connection to a support ring which sitsat the outer rim of the disks This support ring isthen connected to the global support barrel of theMVD (figure 521)

Figure 521 Sketch of the support structure for thestrip disks


Figure 522 Sketch of the common support barrel for BL3 and BL4 with the sawtooth connections for the supermodules The picture at the top shows a first prototype of the support barrel with the sawtooth connectors

Figure 523 Sketch of the set of two strip super modules with a common cooling pipe


56 The Cooling System

The cooling system project provides to operate closeto room temperature with the temperature of thecooling fluid above the dew point The difficultyin retrieval of the most appropriate coolant for therequired operating temperatures and at moderatepressures led to discard evaporative solutions infavor of a depression system using water as cool-ing fluid at 16 C A depression system should avoidleaks and reduce vibrations and stress on plastic fit-tings and manifolds because of moderate pressure

For geometry constraints and installation proce-dure the MVD structure is divided in two halveswhich will be connected together after target pipeinstallation In addition due to the MVD collo-cation relatively to the beam and target pipes allthe services and cooling pipes are routed from theupstream side These constraints have conditionedthe cooling design which always with the view tosaving material has made use of U-shape tubes toavoid aimless material for the return lines and toreduce the number of cooling tubes For an effi-cient cooling system it is advisable to work witha turbulent regime flow which means consideringan inner diameter of 184 mm for the cooling pipesand water as cooling fluid with flow rates higherthan 025 lmin In a depression mode system thepressure drops in the pipes are limited confiningcooling mass flows in reduced adjusting ranges Infact the part provided with the pressure starts justnear the MVD in the inlet pipe till the end of thecooling line (at the cooling plant about 20minus 30 mfar from MVD) Pressure sensors on inlet (and re-turn) lines ensure the depression mode in the MVDvolume

561 The Pixel Cooling System

The pixel volume of the Micro Vertex Detector (seefigure 524 consists of a cylinder of 300 mm in di-ameter and 460 mm in length in which 810 readoutchips dissipate 1 Wcm2 for a total power of about14 KW

The main purpose of the cooling system is to guar-antee the maximum temperature of the assemblyunder 35 C A dry air flow (nitrogen could be bet-ter) has to be foreseen in the MVD volume For thecooling system design a material with low densityhigh thermal conductivity low thermal expansioncoefficient machinable gluable stable at differ-ent temperatures and radiation resistant has beensearched The material which complied to all theserequirements was the carbon foam his open pore

structure graphite produces a material with goodthermal properties and low density In table 52and 53 density and thermal conductivity of twocarbon foams POCO HTC and POCO FOAM arepresented

POCO HTC

Density 09 gcm3

Thermal conductivity

Out of plane 245 W(m middotK)

In plane 70 W(m middotK)

Table 52 POCO HTC properties

POCO FOAM

Density 055 gcm3




Table 53 POCO FOAM properties

Tests have been made on POCO HTC and onPOCO FOAM to verify carbon foam thermal con-ductivity and mechanical stability at the radia-tion levels expected in the experiment Using theTRIGA MARK II reactor in Pavia five samples perfoam type have been immersed in the reactor cen-tral channel at different reactor power for a timeof 1000 minus 1500 s while one sample for foam typenot irradiated was used as reference The testsfor the Youngrsquos modulus estimation consisted on acarbon foam specimens (15times 50 mm2 and 5 mmof thickness) irradiated at different radiation lev-els and then tensioned with a set of loads Forthe evaluation of the deformation strain gaugeshave been glued on the specimens and read by aHBM-MGCplus acquisition system Test resultsare shown in figure 525 and 526 The carbonfoam stiffness increases with the neutron fluencethe POCO FOAM Youngrsquos modulus doubles in theconsidered irradiation range while a 23 variationof the nominal value is observed for the POCOHTC

The tests for the thermal conductivity estima-tion consisted on the same carbon foam specimensheated on one side with a resistance and cooled onthe other side For the evaluation of the thermalconductivity two thermocouples arranged at a wellknown distance have been glued on the specimensand read by the HBM-MGCplus acquisition systemThe results of the thermal tests are reported in fig-ure 527 and 528 the thermal conductivity varia-tions are within 10 for both carbon foams


Figure 524 The Pixel volume with 2 barrels and 6 disks

Figure 525 Test results Youngrsquos modulus [MPa] vsradiation levels for POCO FOAM

Figure 526 Test results Youngrsquos modulus [Mpa] vsradiation levels for POCO HTC

The nominal values pre-irradiation have been usedin the FEM simulations Test results show varia-tion of the thermal conductivity values within 10Because of different geometries different chips con-

Figure 527 Test results thermal conductivity[W(m middotK)] vs radiation levels for POCO FOAM

Figure 528 Test results thermal conductivity[W(m middotK)] vs radiation levels for POCO HTC

centrations different mechanical structure solu-tions the cooling system design can be distin-guished in two sub-groups the disks and the barrels


Disk Cooling System

The cooling system for the disks consists of tubes of2 mm external diameter inserted in a carbon foamdisk with a thickness of 4 mm on which the detec-tors are glued on both sides The tubes with a wallthickness of 80 microm are in MP35N a Nickel-Cobaltchromium-molybdenum alloy (by Minitubes) withexcellent corrosion resistance and good ductilityand toughness In order to save material to re-duce pressure drops and to allow irregular pathsto reach the manifolds of the first patch panel themetal pipes are connected to flexible polyurethanetubes [15] through small plastic fittings (in rytonR4 custom made) The plastic fittings are glued onthe metal cooling tubes with an epoxy glue (LOC-TITE 3425) and all the connections are individu-ally tested to ensure the pressure tightness Thedisks group consists of two disks (4 halves) with adiameter of 75 mm dissipating about 70 W (35 Weach disk) and of four disks (8 halves) with a di-ameter of 150 mm dissipating about 740 W (about185 W each disk) On the small disks due to spaceconstraints only one tube for each half disk is fore-seen while for the bigger disks three U-shapedtubes for each half has been arranged The U-tubeis obtained with plastic elbows (in ryton R4 cus-tom made) glued on the metal cooling pipes Infigure 529 and 530 the two kinds of half disks incarbon foam completed of detector modules cool-ing tubes and plastic fittings are shown

Figure 529 Half small disk (r = 3656 mm) in carbonfoam completed of chips detectors cooling tube andfitting

Finite Element Method (FEM) analyses throughtemperature maps allow fast evaluations of heatconduction identifying the most effective solutions

Figure 530 Half big disk (r = 7396 mm) in carbonfoam completed of chips detectors cooling tube andplastic fittings

for the future prototype constructions After afirst prototype useful to estimate the errors dueto contact surface and to manual gluing manyFEM simulations have been performed to find outthe thermally most efficient configuration of carbonfoam and cooling pipes considering different car-bon foam thicknesses tube diameters tube num-bers and cooling flow rates

From FEM analyses it appears that the most ef-fective operation to be applied is the doubling ofcooling pipes

The test on a second prototype validates FEM re-sults In figures 531 and 532 FEM analyses com-pared to test results are shown

In figure 531 a FEM model with the real configura-tion of half disks in carbon foam with a thicknessof 4 mm six cooling pipes with pixel chips on bothsides and a total power of 94 W was used

In figure 532 a FEM model with the dummy chipslayout simulating the pixel chips dissipating 94 Won the carbon foam disk with a thickness of 4 mmand six cooling pipes was used

For both these FEM analyses cooling water at185 C and flow rate of 03 lmin in each tubehave been adopted Figure 533 shows the disk pro-totype in carbon foam with six cooling tubes insideand 54 dummy chips (resistors)

Figure 534 shows the test results in line with FEMpredictions In this figure some hot spots are visibledue to manual gluing imperfections


Figure 531 FEM analysis results with the follow-ing real configuration total power 94 W removed by6 cooling pipes (2 mm external diameter) cooling flow03 lmin at 185 C carbon foam thickness 4 mm

Figure 532 FEM analysis results with the follow-ing test configuration total power 94 W removed by6 cooling pipes (2 mm external diameter) cooling flow03 lmin at 185 C carbon foam thickness 4 mm

To guarantee a depression system considering thepressure drops for the return lines the pressuredrops in the MVD line between the patch panelin which pressure sensors can take place must belimited and below 450minus 500 mbar

In the disks to reduce the number of cooling cir-cuits it was thought to use S-shaped tubes but thehigh pressure drop did not allow this solution

In figure 535 and 536 the pressure drops in S-tubesand in U-tubes are shown

These measured pressure drops are referred to theonly metal pipe while the pressure sensors will bepositioned in the patch panel manifolds at least 1 mfar from MVD and after plastic tubes connections

Figure 533 Carbon foam disk with 6 inserted coolingpipes and 54 resistors (dummy chips)

Figure 534 Test results Thermal map of disk pro-totype dissipating 94 W

Figure 535 Pressure drop [bar] in a S-shape tube vsmass flow rate [lmin]

In the first case (S-shaped tube) the pressure dropis already close to the limit but all the path throughplastic tubes and the manifolds is neglected There-fore the solution with S-shaped tubes cannot beadopted On the other hand the U-shaped solu-tion thanks to its limited pressure drops allows amore extended range of flow rate settings


Figure 536 Pressure drop [bar] in a U-shape tube vsmass flow rate [lmin]

Barrel Cooling System

The cooling system is designed to drain the totalpower about 75 W on the first layer and 500 W onthe second by 21 MP35N U-tubes (2mm of outerdiameter and 80 microm of thickness) The two halvesof the layer 1 due to cooling constraints in use U-shaped tubes cannot be symmetric one half in-cludes 8 staves and the other only 6

Furthermore the barrel system besides the drain ofthe heat dissipation for the material budget lim-its has to provide mechanical support The systemconsists of the cooling tube inserted and glued ina sandwich of carbon foam and an Omega madeby 3 plies of carbon fiber M55J (properly oriented)The high modulus of this carbon foam ensures therigidity of the structure and at the same time actsas heat bridge for the cooling The super mod-ule structure which consists of the silicon detec-tor completed of chips mechanical structure andcooling system is shown in figure 514

Figure 537 Prototype of two staves of the secondlayer of the barrel with dummy chips (resistors) witha power of about 21 W for each stave

Even for the barrel part FEM analyses have beenuseful to find out the most efficient solutions andtest results are presented below Figure 537 showsthe U-shaped cooling tubes in MP35N with plasticfittings and plastic bends glued to the pipes Thechips are simulated by resistors dissipating about21 W in each stave The cooling water for the test

at 18 C was fed with a flow rate of 03 lminThe temperature map shown in figure 538 wasbeen obtained with a thermo camera (NEC thermotracer)

Figure 538 Test result Thermal map of two stavesprototype dissipating 42 W in total

562 The Strip Cooling System

The strip part of the MVD is subdivided into abarrel section and a disk section The barrel sec-tion again consists of two barrels with 552 front-endchips needed for the inner one and 824 front-endchips for the outer one Compared to the num-bers in table 24 these higher numbers consider thefall back solution of an individual readout for eachsensor instead of a ganging of long strips for sub-sequent barrel sensors The disk section consists ofone double disk with 384 front-end chips Each ofthese front-end chips consumes up to 1 W of powerwhich sums up to a total power consumption of thestrip part of 176 kW

To cool away this power the same concept as in thepixel part is used A water based system will beused which is operated in depression mode Eachsupport structure is equipped with a cooling tubeof 2 mm diameter and a wall thickness of 80 micrommade out of MP35N This cooling tube is glued withthermal conducting glue to a carbon foam material(POCO HTC) with excellent thermal conductivitywhich sits underneath the front-end electronics (seefigure 539 for the strip barrel part)

Detailed finite element simulations of different se-tups have been performed for the barrel part of thestrip detector The results of the best setup can beseen in figure 540 The maximum temperature thefront-end electronics reaches is 32 C in this setupAlmost all heat of the electronics is cooled away

In the strip disk part the situation is more com-plicated because of a higher density of the read-


Figure 539 Schematic view of the cooling concept of the strip barrel part The relevant parts are front-endchip (red) thermal glue (yellow) carbon fiber (green) Rohacell (orange) carbon foam (violet) cooling pipe(black) cooling water (blue)

Figure 540 FEM analysis result for one barrel section of the strip part with 20 strip readout chips on top and 12chips on the bottom side The used parameters are 32 W total power of the chips the chips are glued to carbonfoam with an internal cooling pipe with 2 mm diameter a water flow of 03 lh and 16 C inlet temperature Leftpicture shows the top side of the super module right picture shows the bottom side of the super module

Figure 541 FEM analysis result for one disk sectionof the strip part of the MVD with 8 strip readout chipson the top and 8 chips on the bottom side The usedparameters are 32 W total power of the chips the chipsare glued to carbon foam with an internal cooling pipewith 2 mm diameter a water flow of 03 lh and 16 Cinlet temperature

out electronics Here the achieved temperature ofthe readout electronics would be 375 C whichis slightly above the required 35 C To solve

this problem without changing the cooling conceptwould require to either reduce the power consump-tion of the electronics or double the number of cool-ing pipes which would result in a chip temperatureof 28 C While a doubling of the cooling pipeswould significantly increase the material budget inthis part of the detector the first solution would bemore desirable but it is not clear if this goal can bereached

563 Cooling Plant

The cooling plant will be placed in a reserved anddedicated area at about 20 minus 30 m far from theMVD final position In this area the inlet pump tofeed all the cooling circuits the tank for the waterkept in depression mode by vacuum pump the heatexchanger connected to a chiller the water clean-ing apparatus the control unit and the manifoldsfor the supply and the return lines of the coolingcircuits each equipped with valves regulators andsensors will be located A simple scheme of thehydraulic circuit is shown in figure 542


Figure 542 Simple hydraulic scheme of the cooling plant TT Temperature Transmitter FT Flow TransmitterPreg Pressure Regulator PT Pressure Transmitter

Figure 543 Preliminary draft of the cooling plant


For fault-tolerance reasons the pixel cooling systemis segmented in different cooling loops 9 for the bar-rels and 12 for the disks respectively In the barrel3 cooling circuits are foreseen for layer 1 and 6 forlayer 2 In the disks 4 cooling circuits are for thetwo small disks and 8 cooling circuits for the biggerones The total number of cooling circuits for thepixel part is 21 For the strip part there are 32 cool-ing circuits 24 for the barrel part and 8 for the strippart The most important requirement for the hy-draulic system is that it works at sub-atmosphericpressure all along the MVD ensuring a leak-lesssystem To achieve the sub-atmospheric pressureinside the detectors the pressure drops along thedetectors and the return lines are limited

Pressure sensors on the supply lines at the en-trance of the MVD ensure to operate in depressionmode while flow meters and pressure regulators oneach cooling circuit positioned at the cooling plantguarantee the pressure stability and the nominal op-erating conditions

All the metallic tubes and the wet parts of all theaccessories (sensors valves flow meters filters andfitting) will be in stainless steel AISI 316 The flex-ible tubes will be in polyurethane

The water purity required for corrosion concerns ofthin cooling pipes will be obtained with a cleaningtreatment chain which consists of deioniser microporous membrane to remove the dissolved O2 andsterilisation UV lamp [16]

In figure 543 preliminary drafts of the cooling plantare shown

57 DCS

For a reliable and safe detector operation the moni-toring of operating parameters is crucial Thereforea Detector Control System (DCS) will be deployedto permanently monitor parameter of importancefor the MVD This data will be collected and pro-vided to the shift crew while running the experi-ment The occurrence of critical values outside pre-defined parameter ranges need to create an alert oreven cause an interrupt so that measures can betaken to prevent disturbance in operation or even adamage to the detector The data shall be archivedin order to evaluate the detector performance overtime and identify possible problems The PANDAExperiment Control System (ECS) will be based onEPICS1 It allows to collect display and archivesensor readings from various devices The followingparameters should be kept under surveillance

bull voltage and current of the front-end electronic

bull voltage and leakage current of the sensors

bull temperature of electronic and coolant

bull coolant flow

bull temperature and humidity in the detector vol-ume

The supply voltages and the current consumptionsof the front-end electronics will be measured at thepower supplies and the voltage regulators close tothe detector

For the silicon sensors the depletion voltage and theleakage current have to be supervised They give anindication of the condition of the sensors and can beused to check the radiation damage the sensors havetaken Here a measurement at the power supplieswill be performed

The sensors and to a lesser extend the electronicsare very sensitive to the operation temperature Aclose surveillance of the temperature and the cool-ing system are therefore very important for a safeand stable operation of the MVD

Thus each pixel module and each strip super mod-ule will be equipped with an NTC thermistor tomeasure the temperature on the module The read-out cables of the resistors are routed along with thesignal bus to the patch panel Here the resistivity ismeasured and the data is fed into the DCS system

The cooling system is monitored via temperatureand pressure sensors on each cooling pipe and thecorrect temperature of the water in each pipe canbe kept by a dedicated heater In addition the over-all temperature and humidity in the MVD and thetemperature of the off-detector electronics will bemeasured and controlled

References

[1] P Moreira et al The GBT project Proceedingsof the TWEPP-09 Conference CERN-2009-006342ndash346 2009

[2] J Troska et al The versatile transceiver proofof concept Proceedings of the TWEPP-09Conference CERN-2009-006347ndash351 2009

[3] P Moreira et al The GBT-SerDes ASIC pro-totype JINST 5 C11022 doi 1010881748-0221511C11022

1 Experimental Physics and Industrial Control System


[4] O Cobanoglu P Moreira and F FaccioA radiation tolerant 48 Gbs serializer forthe Giga-bit transceiver Proceedings of theTWEPP-09 Conference CERN-2009-006570ndash574 2009

[5] M Menouni P Gui and P Moreira The GB-TIA a 5 Gbits radiation-hard optical receiverfor the SLHC upgrades Proceedings of theTWEPP-09 Conference CERN-2009-006326ndash330 2009

[6] L Amaral et al A 5 Gbs radiation toler-ant laser driver in CMOS 013 microm technol-ogy Proceedings of the TWEPP-09 Confer-ence CERN-2009-006321ndash325 2009

[7] A Gabrielli et al The GBT-SCA a radia-tion tolerant ASIC for detector control applica-tions in SLHC experiments Proceedings of theTWEPP-09 Conference CERN-2009-006557ndash560 2009

[8] S Baron JP Cachemiche F Marin P Mor-eira and C Soos Implementing the GBT datatransmission protocol in FPGAs Proceedingsof the TWEPP-09 Conference CERN-2009-006631ndash635 2009

[9] B Allongue G Blanchot F Faccio CFuentes S Michelis et al Low noise DC to DCconverters for the SLHC experiments Jinst 52010

[10] D Calvo P De Remigis M Mignone TQuagli and R Wheadon First prototypes oflow mass cables for the pixel detector of theMicro-Vertex-Detector PANDA MVD note09 2011

[11] S Bonacini K Kloukinas and P Mor-eira Elink a radiation hard low powerelectrical link for chip to chip communica-tion TWEPP-09 Proceedings CERN-2009-690422ndash425 2009

[12] GR Almen PD Mackenzie V Malhotra andRK Maskell Fiberite 954 cyanate ester sys-tems SPIE - Design of Optical Instruments1960288ndash294 1992

[13] GR Almen PD Mackenzie V Malhotra andRK Maskell 954 A family of toughenedcyanate ester resins Proceedings 23rd Inter-national SAMPE Technical Conference pages947ndash958 Oct 21-24 1991

[14] G Giraudo et al The SDD and SSD supportstructure for the ALICE Inner Tracking Sys-tem Jinst 4 P01003 2009

[15] M Tavlet A Fontaine and H SchonbacherCompilation of Radiation Damage Test DataThermoset and thermoplastic resins compos-ite material CERN 98-01 Technical InspectionAnd Safety Commission May 1998

[16] M Pimenta dos Santos Report EngeneeringSpecification EDMS 502

103

6 Monte-Carlo Simulations and Performance

A full simulation of the PANDA detector compo-nents and their answer to the passage of particleshas been developed within the PandaRoot frame-work [1] The simulations for the MVD are imple-mented within that framework The same recon-struction and analysis algorithms and infrastruc-ture will work on the simulation output as well ason real data

This chapter will give an overview of the frameworkthe MVD detector and reconstruction code as wellas the results of performance studies

61 Software Framework Layout

The software framework is the collection of soft-ware and tools for the description of the detectorsand the simulation of physics reactions The lay-out is organized such as to allow the re-use of wellknown programs and tools common to particle andnuclear physics simulations One basic concept ofthe framework is its modularity the possibility toswitch algorithms and procedures at mostly anypoint in the chain of processes Hence a set of inter-faces and data inputoutput is provided to connectall these tools properly

In practice the frameworks organization featuresthese levels

bull External Packages containing the main bulk ofsoftware Geant3 and Geant4 (Particle prop-agation through matter in Fortran [2] andC++ [3]) VMC (Virtual Monte-Carlo [4] [5])ROOT (Plotting fitting graphics etc [6])Pythia [7] and auxiliary tools

bull FairRoot handling the framework data IOinterfaces infrastructure

bull PandaRoot for detector simulations trackingand reconstruction

To this framework user processes are added asdetector class implementations for the transportmodel and as tasks which steer the processing atany stage The availability of online parameters isprovided with the runtime database (RTDB) anddata are stored in a ROOT file using ROOTs ob-jects handling with chains trees and branches Us-ing VMC together with the geometry description in

Figure 61 Schematic chain of processing

ROOT format it is possible to switch between dif-ferent transport engines This enables an easy wayto compare the outputs of different engines witheach other and with real data in order to get a re-liable description of the detectors behaviour

A typical chain of processing is sketched in fig-ure 61 It has the following stages

Event Generator The event generator producesin each event a set of particles to be processed Par-ticles are defined by their four-momentum chargecreation point and time These properties are ran-domly distributed by the selected model such as thetechnically important ldquoparticle gunrdquo single-channelgeneration with EvtGen [8] or the full descriptionof antiproton-proton or antiproton-nucleus back-ground reactions with DPM1 and UrQMD2 respec-tively Also Pythia and Pluto are available

Transport Code (VMC) The particle transportthrough matter is usually simulated with one ver-sion of Geant The Virtual Monte-Carlo interface(VMC) allows to switch between the different Geantflavors and allows to introduce other packages aswell Each sub-detector has its own geometry de-scription (ROOT format) and the and the defini-tion of the active elements in which the interac-tion of particles can occur according to differentselectable physical processes (such as energy lossby ionization showering or Cherenkov emission)

Digitization All Monte-Carlo-Data are processedto model the detectors answer or to produce theeffective answer of a readout chip The digitizeddata formats out of the simulation must be the sameas those obtained in a real implementation such asa prototype detector or even the final setup

Local Reconstruction The digitized data aresubject to a local reconstruction procedure which

1 Dual Parton Model [9]

2 UltraRelativistic Quantum molecular Dynamics modelfor pN [10][11]


Figure 62 Basic approach for the introduction ofCAD models into physics simulations [18]

associates them to information with a physicalmeaning like a 3D point (from MVD GEM andcentral tracker) a total energy (from EMC) or aCherenkov angle (from DIRC RICH)

Tracking The tracking procedure in PandaRootproceeds through two steps track finding (pat-tern recognition and track prefits) and track fitting(Kalman filter) While the track finder algorithmsare very different for each sub detector (or combina-tions of them) the fitting is done by means of theGENFIT package [12] based on the kalman filtertechnique [13] [14] using GEANE ([15] [16] and[17]) as propagation tool

PID Particle Identification is performed globallytaking information from all detectors into accountEach detector provides probability density func-tions (pdf) for each particle kind merged to-gether to provide a global identification probabilityusing Bayes theorem In addition different multi-variate algorithms are available such as a KNN3

Classifier or a Neural Network approach The tun-ing of these algorithms still is ongoing

Physics Analysis The analysis tools have todeal with a collection of information based on four-momenta positions and the identity of the recon-structed particles in a unified way Particle com-bination selection mechanisms and manipulationtools (like boosting between lab and center of massframe) are provided Furthermore a set of fitters isavailable to fit the four momenta and positions ofparticles (eg from a decay) under different typesof constraints

62 Detector Model of theMVD

Along with the hardware developments a very de-tailed CAD model of the MVD has been accom-plished which delivers a very comprehensive andaccurate detector description [18] It is based on a

design optimization between a good spatial cover-age on one hand and the introduction of sufficientspace needed for passive elements and the detec-tor integration on the other Thereby a minimummaterial budget is achieved The model includes ac-tive silicon detectors connected readout electronicslocal and global support structures cables a dedi-cated cooling system as well as the routing of all ser-vices In addition connectors and electronic com-ponents such as surface-mounted devices (SMD) areintroduced In total this comprehensive detectordescription contains approximately 70500 individ-ual volumes

A converter [19] has been developed to transformthe CAD model into a ROOT geometry suited forsimulations within the PandaRoot framework Ituses a common output format of CAD software(STEP format [20]) as input data Besides theproper definition of the entire CAD in a ROOTgeometry material properties ie the density andthe radiation length are given explicitly for each ofthe components Therefore basic conventions havebeen fixed which allow for an unambiguous andautomated allocation of each single volume [21] Aschematic illustration of the procedure described isgiven in figure 62

The main structure of the available detector geom-etry is shown in figure 63 At the top an illustra-tion of the model before and after the conversionis shown Basically there are four functional partswhich are introduced as independent objects Thesilicon part contains all silicon detectors ie thepixel and the strip sensors along with all associ-ated readout chips The second group is formed bythe support structures of the detector It includesthe local support for detector modules and the indi-vidual detector layers as well as the global frame forthe overall detector integration of the MVD More-over a schematical support structure needed for therouting of services in upstream direction is addedAll support components are defined as lightweightcarbon composites of different effective density andradiation length

The entire model of the routing part scales withthe number of readout channels and is based onlow-mass cables and the overall cooling concept out-lined in chapter 54 and 56 respectively Differentcable types are introduced for the high voltage sup-ply of the sensors the power supply of the read-out electronics as well the data transmission andslow control of the readout electronics All cablesare split into a conductive core and an insulationlayer The cooling lines consist of a water core and

3 K Nearest Neighbors Classifier


Figure 63 Top View of the detailed MVD model before (a) and after (b) the conversion from a technical CADproject into a ROOT geometry The main structure of the model is shown at the bottom

a casing material which is defined either as steel orPVC The barrel layer services are guided straight-forward in upstream direction following the shapeof the target pipe The disk services are lead outto the top from where they are further extended inupstream direction The concept for the pixel disksis more complex because the inner layers are in-serted inside the barrel part Therefore they mustbe routed at first in forward direction until a posi-tion from where they can be brought to outer radiiWhile for all other parts a circular arrangement ischosen the routing of the pixel disks is performedat the top and bottom

The last part of the MVD model includes addi-tional components of the electronics and coolingsystem In the latter case these are given by con-nectors needed for the intersection from steel pipesused on the detector modules to flexible pipes atthe outer part of the cooling circuits The electron-ics part includes capacitors and SMD componentsie condensators and resistors in order to delivera complete description of fully equipped detectormodules

63 Silicon Detector Software

A software package for strip and pixel silicon sen-sors called Silicon Detector Software (SDS) hasbeen developed to reproduce both the MVD andthe Luminosity monitor in the PandaRoot frame-work

631 Monte-Carlo Particle Transport

Particles produced in the event generation phaseare transported through the detector geometry andthe magnetic field by the selected Monte-Carlo en-gine When a particle crosses an active volume aMonte-Carlo object is stored containing the entryand exit coordinates the total energy loss inside thevolume as well as the exact time the interaction inthe detector occurs

Passive detector components eg holding struc-tures contribute with their material budget to scat-tering energy loss secondary particle production(delta-electrons) etc

632 Digitization

Digitization is the emulation of the detector an-swer before using reconstruction algorithms The


x

Q

chargecloud

particle track

sensor

Figure 64 Schematic sensor profile (asymp 100minus 200 micromthick) view and the linear digitization model with acharge cloud Below the charge distribution on the read-out cells is illustrated

entry and exit coordinates and the charge informa-tion are used to determine which pixels or stripsfired The output data is the same format like thereal data will have Each fired strip or pixel is calleda ldquodigirdquo identifying the strippixel itself as well asthe involved front-end chip and the sensor Fur-thermore the measured charge and the determinedtime stamp are stored The digitization is a processin two stages geometric and electronic digitizationdescribed below

Geometric Digitization

Geometrically one can project the particle paththrough the sensor to its surface As the pathcrosses the readout structure the discretization intopixels or strips is done The charge collected in eachof the channels is calculated by a gaussian cloud dis-tribution around the path (illustrated in figure 64)The charge Qi for one strip is given by

Qi =Q

|xout minus xin|middot (Fi(xout)minus Fi(xin)minus Fi+1(xout) + Fi+1(xin))

(61)and the charge Qij in one pixel is

Qij =Q

4 middot |xout minus xin| middot |yout minus yin|middot (Fi(xout)minus Fi(xin)minus Fi+1(xout) + Fi+1(xin))

middot (Fj(yout)minus Fj(yin)minus Fj+1(yout) + Fj+1(yin))(62)

Q being the total number of electrons produced bythe particle xin and xout (y respectively) the entry

Figure 65 Simple model of the signal shape at theamplifier output

and exit coordinates of its track and Fk(x) being

Fk(x) = (xkminusx) middoterf

(xk minus xradic

2σC

)+

radic2σ2

C

πmiddoteminus (xkminusx)2

2σ2C

(63)The width σC of this cloud is a parameter which isassumed to be in the order of 8 microm

Electronic Digitization

Charge from the sensors is measured in thefront-end electronics with the Time-over-Thresholdmethod (ToT) (See also section 333) The chargeis integrated in a charge sensitive amplifier (CSA)at the same time the capacitor of the CSA is dis-charged by a constant current source The outputof the CSA then is compared with a threshold volt-age The time the amplifier output is above thisthreshold is proportional to the deposited charge

The SDS package provides the possibility to includedifferent models for the ToT calculation

The simplest model for the front-end chip behavioris the triangular one (See figure 65) It assumes alinear rising and falling of the signal with a constantpeaking time The characteristic parameters are thepeaking time the constant discharge current andthe discriminator threshold

For the event building it is necessary to mark everydigi with precise time information the Time StampIn reality the Time Stamp is set by the front-endelectronics when the rising signal crosses the thresh-old voltage Due to the not negligible rising timeof the preamplifier this time is not the same instantthe particle hits the sensor This delay between thephysical hit and the recognition by the electronicsis called Time Walk (figure 65)

The simulated Time Stamp of a digi is the sum ofthe Event Time (MC information) the flight timeof the particle till it hits the frontside of the sensorand the Time Walk of the Signal


Q0 1000 2000 3000 4000 5000 6000

-710

-610

-510

-410

-310

-210

-110

1Threshold

NoiseDistribution

Probability ofNoise Hit

Figure 66 Illustration of the probability for a pixelor strip to produce a noise hit (grey area)

633 Noise Emulation

Electronic noise has the character of a gaussian dis-tribution centered around the baseline with thewidth of σnoise The noise baseline is set as thenominal zero in the simulations To deal with theelectronics noise each collected charge in a pixel andstrip channel is redistributed by the noise gaussianwith σnoise = 200 electrons for pixel sensors andσnoise = 1000 electrons for the strip sensors asthe hardware measurements show (see sections 33and 42) A threshold cut is applied to suppress toolow charge entries being set to 5σnoise The gaus-sian tail over the threshold will produce hits by thenoise statistics (see figure 66) This is taken intoaccount by adding randomly selected channels Thenumber of noise hits is calculated by a poissonianrandom distribution with its mean value being

Nnoise = Nchannels middot tevt middot fclock middot1

2erfc

(Qthrradic

2 middot σnoise

)(64)

while tevt is the time interval from the previousevent and fclock the bus readout frequency anderfc(x) being the complementary error function

634 Local Reconstruction

Digitized data are reconstructed by finding clustersand forming hits at the clusters centeroid using thecharge measurements to improve the accuracy Fur-thermore the time of the particle passage is recon-structed by the time walk correction

Cluster Finding

Cluster finding means to collect all digis belong-ing to one track passing through the sensor Thisis done iteratively by taking all signaling channelsaround one digi wthin a given tolerance of channelsA minimum charge sum of a clusters digis can berequired to reject ghost hits produced by noise only

Hit Reconstruction

Having a cluster of digis calculating the centeroidfinds several approaches [22]

Simply taking the central position of the channelwith the highest charge yield in a cluster defines thebinary centeroid Its uncertainty is ∆x = p

radic12

This method is the only option to calculate coordi-nates with only one entry per cluster

A better resolution for clusters with two or moredigis is given by the center of gravity which uses thedigis charge measurements as weight The averageposition is calculated by

x =

sumi qixisumi qi

(65)

This approach mirrors the geometric digitizationprocedure in the sense of isotropy in the readoutstructures and the lack of magnetic field effectsThe precision ∆x is smaller than p

radic12 which will

be shown in section 67

To get a precise 3D information with the strip sen-sors top and bottom side hits have to be corre-lated This produces ambiguities when at least oneof the sides of a sensor has more then one clusterSuch a situation is created by multiple tracks hit-ting a sensor in one event or in case of hits generatedby electronic noise Because the charge carriers ar-riving at either side correspond to pairs producedby the same particle a correlation in the measuredcharge (in form of a cut in the charge difference|Qtop minusQbottom| lt Qcut) can reduce the number ofambiguities

Time Walk Correction

To improve the time resolution of the TimeStampa correction of the TimeWalk effect can be appliedBased on a ToT conversion model the TimeWalkcan be calculated from the ToT value and sub-tracted from the TimeStamp


Figure 67 Track finding efficiency of the Riemann-Track-Finder for six pion tracks with random momen-tum and theta angle in the MVD vs the momentumof a single track The error bars indicate the range ofefficiency for different theta angles

For the triangular ToT model the TimeWalk isgiven by

tTW =QThr[e

minus]

Q[eminus]middot trise (66)

64 Tracking and Vertexing

The tracking procedure in PandaRoot proceedsthrough three steps First local pattern recogni-tion algorithms find tracklets in different subdetec-tors Afterwards a global track finding and trackletmerging is performed The global tracks are thenglobally fitted

641 MVD Tracklet Finding andFitting

The track finding and pre-fitting for the hit pointsinside the MVD is based on a fast circle fit usingthe x-y-coordinate of a hit point and a second linearfit of the arc length of a tracklet on the fitted cir-cle and the z-coordinate For a computational fastcircle fitting a translation of the hit points to a Rie-mann surface is performed Here a plane is fittedto the track points The parameters of the planecan be than back calculated to the circle parame-ters mid point and radius A detailed descriptionof the algorithm can be found in [23]

The performance of the track finder is shown in thefollowing figures Six charged tracks per event witha random angular distribution and random momen-tum between 01 GeVc and 1 GeVc have been sim-ulated In the first diagram (figure 67) the number

Figure 68 Track finding efficiency for tracks with atleast four hits in the MVD of the Riemann-Track-Finderfor six pion tracks with random momentum and thetaangle vs the momentum of a single track The errorbars indicate the range of efficiency for different thetaangles

of correctly found tracks divided by the number ofall simulated tracks is plotted against the momen-tum of a single track The error bars indicate thevariation of the track efficiency for different thetaangles For tracks with very low momenta of about100 MeVc the finding efficiency is about 77 whichrises up to 88 for tracks above 600 MeVc Theincrease of track finding efficiency with rising mo-mentum can be explained by the dominance of mul-tiple scattering for low momentum tracks The lowoverall track finding performance inside the MVDis coming from the low number of hit points pertrack In the second diagram (figure 68) the ef-ficiency was calculated for all tracks with at leastfour hit points in the MVD which are required fora successful track finding

If one calculates the tracking efficiency only forthose tracks with at least four track points it startswith 88 at 100 MeVc and rises up to 97 fortracks above 400 MeVc The error bars are exceed-ing 100 due to tracks originating from the pri-mary vertex with only three hit points They canbe found by the Riemann-Track-Finder as well butthey are not used for the efficiency evaluation

To improve the overall track finding efficiency acombination of the MVD hit points with those ofthe central tracker is necessary as explained in sub-section 642

642 Track Finding

The global track finding operates in a sequenceeach step combining several detection systems


First tracklets from the STT pattern recognitionare merged with the MVD tracklets and unmatchedMVD hits Second any unmatched MVD track-lets are merged with remaining STT hits Thenall remaining MVD and STT hits are checked forforming new track candidates Finally the GEMdisk hits are matched to the existing track candi-dates Each step features an update of the trackdescription parameters as well as a more stringentcondition to clean the candidates of spurious solu-tions A detailed description can be found in theSTT technical design report [24]

643 Track Fitting

Track fitting in PANDA is performed globally usingall tracking devices It employs the Kalman Filter([13] [14]) technique using the GENFIT package[12] Particle transport is done with the GEANEpackage ([15] [16] and [17]) which backtrackscharged particles in a magnetic field taking careof energy loss and scattering in materials For eachtrack a momentum resolution of ∆pp lt 2 (20 ltΘ lt 140) and ∆pp lt 3 (12 lt Θ lt 20) isachieved [24] By default all tracks are propagatedto the point of closest approach to the z axis to formparticle candidates

644 Vertex Finding Algorithms

Finding the common starting point (vertex) of sev-eral tracks is the central goal of the MVD Thereare several approaches under development For pri-mary vertices and short-lived particle decays alltracks are treated purely as helices because scatter-ing and energy loss are negligible inside the beampipe vacuum

POCA Finder

A first approach is to find the point of closest ap-proach (POCA) of two tracks which is done analyt-ically Both helices are projected to the x-y-planeassuming a constant magnetic field in z directionforming circular trajectories The intersection orminimal distance of the two circles defines the de-sired region This is not the POCA in a full 3Dtreatment but produces a good approximation fortracks originating from the same vertex The calcu-lated distance between the tracks gives a measureon how well this assumption holds

Having more than two tracks combined the ideais to average the POCAs of each combination of

two tracks weighted by the inverse of the tracksdistances Dij For n tracks one gets the vertex ~V

~V =

sumni=0j=i+1D

minus1ij

~Vijsumni=0j=i+1D

minus1ij

(67)

To evaluate the rdquogoodnessrdquo of the retrieved vertexthe inverse of the weights sum is used

D =

nsumi=0j=i+1

Dminus1ij

minus1

(68)

This method does not take the tracks covariancesand error estimates into account hence is not ableto estimate the vertex covariance and error How-ever in order to find a seed value for a fitter or toreject outlying tracks from bad combinations thisansatz is fast and useful

Linearised Vertex Fitter

In this approach [25] the tracks are locallyparametrized and their propagation close to the ex-pansion point4 is linearized The fitrsquos basic ideais to reduce the dimensions of the matrices to beinverted For n tracks one gets n 3 times 3-matrices(5 being the number of parameter necessary to de-scribe a helix while two of them the polar angleand the trajectory curvature remain fixed) to beinverted instead of one 5ntimes5n-matrix in the generalcase This improves the calculation speed of the fit-ter but lacks precision The procedure is slightlysensitive to the choice of the expansion point thusa good estimate on the vertex position is required

Because of the imposed linearisations the calcula-tion of the track parameters and covariances featuremainly sums over the tracks Hence it is quite easyand fast to add or subtract tracks from the set with-out repeating the full calculations This procedureis very useful to reject outlying tracks or to formmother particle candidates ldquoon the flyrdquo by vertexconstraint only

The procedure may be separated into two partsone being a fast vertex fitter and the other the fullfitter which also fits the particles four-momentaThe calculation of the covariance matrices and theχ2 gives the estimate of the fit quality

4 Point chosen where to do the parameterization and lin-earization procedure It is the Origin for vertices close to theinteraction point


Kinematic Vertex Fitter

A vertex constrained kinematic fitter is availableIt is based on the Paul Avery papers (eg [26] and[27]) and fits the whole set of momenta and the ver-tex position It employs the inversion of the whole5ntimes5n-matrix for the 5 helix parameters of the ntracks The fit quality is estimated by the χ2 valueand the covariance matrices

65 PID Algorithm for the MVD

The passage of a particle through matter is charac-terised by a deceleration due to the scattering withthe quasi-free electrons therein The energy loss in-formation can contribute to the global particle iden-tification (PID) especially in the low momentumregion (below a MIP) where the Bethe-Bloch dis-tribution is steeper depending on the particle massBecause the energy loss process has a statistical na-ture itrsquos necessary to reproduce the behavior of aprobability density function (pdf) of the energyloss that will depend on the particle momentumand on the measurement uncertainty Different sta-tistical approaches have been investigated in orderto obtain a robust PID algorithm

651 Energy Loss Information

The energy loss information of different particlespecies have been studied and parametrised Theinteraction of various particles has been simulatedwithin PandaRoot using Geant4 and the recon-struction tools available in the framework In orderto take the uncertainty of the detector and of thefull track fit into account the reconstructed trackshave been used The total energy loss of a trackis considered which corresponds to the energy lossthrough few hundreds microm of silicon because thenumber of contributing hits will be at maximum 6in the forward part and 4 in the barrel part Theenergy loss distribution of the reconstructed tracksfor different particles is shown in figure 69

652 Statistical Approach for theEnergy Loss Parametrisation

Different statistical approaches have been studiedto assign a dEdx estimator to each track Thearithmetic mean value of the energy loss of courseis not a good estimator due to the long tail of theenergy loss distribution Several estimators havebeen compared applying different algorithms to

Figure 69 dEdx in the MVD as a function of themomentum

the single-particle tracks To estimate the qualityof the different estimators a comparison in terms ofstandard deviation and of efficiency and purity ofthe samples selected by different algorithms hasbeen performedEach particle species has been simulated at sevendifferent momenta from 400 MeVc up to 1 GeVcin step of 100 MeVc then the energy loss of thereconstructed tracks has been studied The particleenergy lost in each layer dEdx is approximatedby the quantity ∆E(D cos θ) where ∆E isthe energy lost in each sensor (measured by theintegrated charge of a ldquoclusterrdquo ie a set of neigh-bouring pixels or strips) D is the active sensorthickness and θ is the angle between the trackdirection and the axis normal to the sensor surfaceThe distribution of the track energy loss has beenfitted with a Gaussian or with a Landau function inorder to find the mean value or the most probablevalue (mpv) and the corresponding sigma for thedifferent particle species and for different momenta

Mpv of a Landau DistributionA unique specific energy loss ∆E∆x value alongthe whole trajectory is estimated because the lossof energy through each layer is negligible with re-spect to the considered momenta Such a value iscalculated as

∆E∆x =

sumnk=1 ∆Eksumnk=1 ∆xk

(69)

where n is the number of track hits in the MVD∆Ek is the energy lost by the particle in each layer


Figure 610 Separation power as a function of the momentum (a) SP for pK (b) SP for pπ (c) SP for Kπ

and ∆xk is the crossed thicknessA Landau distribution has been used to fit the totalreconstructed energy loss of each track The mostprobable values and the related sigmas have beenused for the parametrisation

MedianThe median of a distribution is calculated by ar-ranging all the ∆E∆x of the track on the differentdetector planes from the lowest to the highest valueand picking out the middle one (or the mean of thetwo middle value in case of even observations) Themean value and the sigma of the Gaussian distri-bution obtained from the different tracks is used asreference value of the energy lost by that particleat a given momentum

Truncated MeanThe truncated mean method discards the highestvalue in a fixed fraction of the total number of ob-servations calculating the mean of the remaindervalues In this study 20 of the hits are discardedThis cut is applied only if a track has more then 3hits

Generalised MeanThe ldquogeneralised meanrdquo of grade k of a variable xis defined as

Mk(x1 x2 xn) =

(1

n

nsumi=1

xki

) 1k

(610)

In this case the grade k = -2 [28] has been used

The separation power (SP [29]) has been calculatedas

SP =|lt dEdx gtp1 minus lt dEdx gtp2|

σp12 + σp22 (611)

where lt dEdx gt is the mean value (or the mpv)of the energy lost by a particle at a certain mo-mentum p1 and p2 are two particle species to be

separated and σ is the standard deviation of the dis-tribution The separation power pK pπ and Kπfor the different algorithms is shown in figure 610As expected the separation power decreases withthe increasing momentum

Even if the Landau distribution shows a better sep-aration power (because of the smaller sigma of theLandau distribution at least by a factor 2) it isnot the most suitable estimator because of the longtail which introduces a significant contamination

In order to build a probability density function thetrend of the mean (or MPV) values and of the sigmaas a function of the momentum have been fitted bythe following function

dE

dx=

a

p2

[b middot ln p2 minus p2 minus c

](612)

where a b and c are the parameters of the fit Asan example the parametrisation of the mean valueof the median as a function of the momentum isshown in figure 611

Figure 611 Fit function of the mean values of themedian distribution as a function of the momentum fordifferent particle species


Figure 612 Efficiency and contamination for protons (left) and kaons (right)

In this way the momentum and the energy lossof a given reconstructed track are used to buildthe probability density function and to extract theprobability to belong to a certain particle species

The efficiency and the contamination of the differ-ent algorithms has been plotted as a function of themomentum for the proton and kaon hypothesis asshown in figure 612 using a probability thresholdof 90 (ie the particle hyphotesis should have aprobability bigger than 90 to belong to a certainparticle species) The efficiency of proton identifi-cation is almost constant around 90 and the con-tamination in the worst case (at larger momenta)is smaller than 30 The efficiency of kaon identifi-cation is more dependent on the momentum but itis larger than 80 at least up to 500 MeVc Thecontamination is higher than in the case of protonby a factor of 2 at 600 MeVc The comparison ofthe three algorithms shows that the efficiency is al-most the same in the three cases while the contam-ination is smaller for the generalised mean Otherparticle species are not taken into account becauseas shown in figure 69 the signals of pions muonsand electrons in the MVD are too overlapped to de-liver a good identification efficiencyThe MVD can contribute to the global PID decisionup to a momentum of 500 MeVc for kaons and upto 1 GeVc for protons

66 General Simulation Studies

661 Radiation Damage

Most important for the damage of the silicon sen-sors is the non-ionising energy loss which causesdamages in the lattice structure of the semicon-ductor material which influences the collection effi-ciency and the leakage current of the sensor Thedamage caused by a particle strongly depends onthe particle type and its momentum To comparethe damage caused by different particles one can de-fine a hardening factor κ which scales the differentparticle types to the damage caused by the flux of1 MeV neutrons according to the NIEL hypotheses[30] This scaling factor can be seen in figure 613for various particle types

The simulation of the radiation damage was per-formed with the PandaRoot framework Dependingon the target material two different event genera-tors have been used DPM1 for hydrogen targetsand UrQMDSmm for nuclear targets The gener-ated particles then were propagated through thefull PANDA detector using Geant4 to include alsobackscattered particles from detectors further out-side of PANDA The neutron equivalent flux of par-ticles then was determined by weighting each in-tersection of a particle with a silicon sensor withthe momentum and particle type dependent scal-ing factor This simulation was done for differenttarget materials and beam momenta from 2 GeVcup to 15 GeVc For each setting more than 106

1 Including elastic Coulomb scattering with a cut off be-low a scattering angle of 017


10-10 10-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 100 101 102 103 104

E[MeV]

10-5

10-4

10-3

10-2

10-1

100

101

102

103

104

D(E

)95

MeV

mb

neutrons Griffin + Konobeyev + Huhtinen

neutrons

protons Summers + Huhtinen

protons

pions Huhtinen

pions

electrons Summers

electrons

Displacement damage in Siliconfor neutrons protons pions and electrons

A Vasilescu amp G Lindstroem

Figure 613 Damage factor for different particle typesand momenta [31]

antiproton-target interactions have been simulatedon the pandaGRID

With the simulated events no time information isgiven Therefore it is necessary to estimate the in-teraction rate R for different target materials andbeam momenta to come from a fluence for a num-ber of events A to a fluence for one operation yearwith 107s

t =A

R=

A

σ middot L(613)

The cross section σ for hydrogen targets is given bythe DPM generator the values for nuclear targetswas taken from [32]

Figure 614 shows the different dose distributionsfor a light hydrogen target and a heavy xenon tar-get for a beam momentum of 15 GeVc The samediagrams can be seen for the pixel part in figure 615for two beam momenta of 2 GeVc and 15 GeVc andfor a hydrogen and a xenon target The colour scaleof the diagrams is different for the different settingand does not allow a comparison between the dif-ferent drawings For the light target the maximumradiation dose is seen for the inner parts of the for-ward disks while for the heavy target the distribu-tion of the radiation dose is more isotropic with amaximum at the innermost barrel layers which areclosest to the target

The two histograms 616 show the overall dose dis-tribution for the two different target materials forone year of PANDA operation One can see a stronganisotropy of the distributions with few very hot re-gions and most areas with more than a magnitudeless radiation damage than the maximum

Figure 614 Relative Dose distribution for a beammomentum of 15 GeVc in the strip part with a hydrogentarget (top) and a xenon target (bottom) [33]

(a) Hydrogen target momentum 2 GeVc

(b) Hydrogen target momentum 15 GeVc

(c) Xenon target momentum 2 GeVc

(d) Xenon target momentum 15 GeVc

Figure 615 Relative dose distribution for a beammomentum of 2 GeVc and 15 GeVc in the pixel partwith a hydrogen and a xenon target From left to rightpixel barrels small disks and large disks [33]


Figure 616 Dose distribution per mm2 and operationyear for a hydrogen target (top) and a xenon target(bottom) with 15 GeVc beam momentum [33]

0

50

100

150

200

250

300

350

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

10

9N

EQ(

mm

2 middot1

07 s

)

p [GeVc]

Xe

Xe_max

N

N_max

p

p_max

Figure 617 Maximum radiation dose per mm2 andoperation year for various target materials for the fullMVD [33]

A summary of the radiation damage for one oper-ation year for various target materials and beammomenta is shown in figure 617 This diagramshows the maximum radiation dose a square mmof sensor material sees in one operation year For ahydrogen target it stays below 1middot1013 n1MeV eq cmminus2

with only a slight increase with the beam momen-tum while for nuclear targets a strong momentumdependency can be seen with a maximum dose rate

of 5 middot 1013 n1MeV eq cmminus2 for a xenon target with15 GeVc beam momentum A duty cycle of 50was assumed for these estimates

662 Detector Coverage

In order to achieve the required tracking perfor-mance a high spatial coverage with a sufficientnumber of track points inside the MVD is neededThe implemented optimisation of the detector ge-ometry aims at a minimum of four MVD hit pointsfor central tracks at polar angles above 10 In thisway an independent pre-fit of MVD tracklets canbe performed by means of fast tracking algorithms(for details see 641) Moreover for a precise vertexreconstruction the first hit point must be detectedas close as possible to the origin of the track

Extensive studies have been performed to evaluatethe detector coverage obtained after the optimisa-tion process Therefore a reduced model of theMVD was used which only includes active sensorelements and no passive material Due to the focuson mere geometrical aspects connected to the sensordimensions and their arrangement in the individ-ual layers all pipes of the beam-target system wereomitted thus minimising scattering effects Mate-rial effects will be explicitly discussed in the nextsubsection 663

In the presented coverage study particles of fixedmomentum were generated at the nominal interac-tion point and isotropically emitted over the po-lar angle θ and the azimuthal angle φ For thefollowing particle propagation through the detectorthe magnetic field of the 2 T solenoid was switchedon Simulations were carried out in the PandaRootframework [34] which allows an interface to theGeant4 transport code [3] Table 61 summarises allbasic input parameters The different setups werechosen in order to check a possible influence of theresulting bending radius on the effective detectorcoverage

MVD model Only active elements

Magnetic field | ~B| = Bz = 2 T

Particle species π+ πminus p p

Momentum [GeVc ] 015 05 10 15

Particle distribution Isotropic in θ and φ

Start vertex (x y z) = (0 0 0)

Number of events 2 million scan

Table 61 Basic parameters for the simulations


Figure 618 Overall detector coverage for pions with amomentum of 1 GeVc emitted from the nominal IP (a)and respective contributions of the pixel and the barrelpart (b)

Due to the high statistical sample for each of thesimulation runs a fine-meshed mapping of the de-tector coverage becomes possible A corresponding2D diagram for one selected configuration is shownin figure 618 (a) In the chosen representation thenumber of obtained MVD hit points per track isplotted colour-coded against its initial emission an-gles θ and φ Contributions of the MVD can beobserved in a polar angle interval from 3 to 150which is in accordance with the basic detector lay-out (cf figure 24) The envisaged design goal ofat least four hit points is achieved in a large rangeof the solid angle Regions with more than fourhit points are related to the radial overlap of ad-

Figure 619 Comparison of the detector coveragefor three different particle species with low momentum(p=150 MeVc)

jacent elements within one barrel layer and inter-leaves of different disk layers in the forward direc-tion The two gaps around θ = 90 reflect the keep-outs needed for the target pipe Larger areas with areduced number of hit points around the target pipegaps and in the transition region from the barrel tothe disk part are related to safety margins requiredfor the overall detector integration They reflect theadjustment to the crossing target pipe in each of thebarrel layers and the fitting of the inner pixel disksinside the MVD barrel part respectively Smallerstructures with a reduced number of hit points re-sult from passive sensor edges

In figure 618 (b) the overall hit distribution is splitinto the pixel and the strip part of the MVD Dueto the rotational symmetry of the MVD halves inφ only the region between 0 le φ le 180 is shownfor each of them In both cases the barrel regionbetween θ = 70 and θ = 140 is covered very ho-mogeneously Larger areas with a reduced num-ber of hit points result from the adjustment to thecrossing target pipe in the different barrel layers Asignificant difference between the two main detectorparts occurs in the forward region (θ lt 60) Thesix double-sided pixel disks deliver a more complexpattern compared to the homogeneous coverage ofthe two strip disks Due to the approximation ofdisk layers with rectangular-shaped pixel sensorsthe covered area in the θ-φ diagram corresponds toa circular arc shaped

A comparison between different configurations isshown in figure 619 The opposed bending radiusof charge conjugated particles with low momenta isreflected by the different curvatures of the overlapsin the barrel part A comparison between differentconfigurations is shown in figure 619 The opposedbending radius of charge conjugated particles with


Figure 620 Yield of different hit counts in four de-tector regions Results obtained for charged pions withdifferent momenta (see table 62) Error bars includedeviations of the different setups and statistical fluctu-ations

low momenta is reflected by the different curvaturesof the overlaps in the barrel part In case of lowmomentum nucleons strong attenuation effects canbe observed They are caused by interactions withthe active silicon material during the particle prop-agation Considering that in a realistic scenario allpipes of the beam-target system and the full mate-rial budget for the MVD must be included it canbe deduced that most of the slow nucleons with mo-menta below 200 MeVc get stuck in one of the MVDlayers In particular this fact becomes relevant forelastically scattered recoil protons

Quantitative results on the relative detector cov-erage in the maximum polar angle range from 3

to 150 are summarised in table 62 They are ex-tracted from simulations with positively and neg-atively charged pions in a momentum range from150 MeVc to 15 GeVc The impact of opposedbending radii for the π+ and the πminus is very smalloccurring deviations are dominated by systematicerrors and statistical fluctuations of the simulationObtained values for a minimum of one and four hitpoints are in the order of 95 and 60 respec-tively The missing 5 in the overall acceptance isrelated to the target pipe gaps

For a more precise evaluation in terms of physicsperformance the total detector acceptance can besplit into four main regions according to the basicdesign of the detector (cf figure 24)

Region I Polar angles between 3 and 10 rep-resenting the solid angle covered by the innerrims of the pixel disks

Region II Polar angles between 10 and 40 rep-resenting the solid angle covered by the forwarddisk layers

Region III Polar angles between 40 and 70 rep-resenting the solid angle covered by the stripbarrel part the outer pixel barrel layer and theinnermost pixel disk delivering the fourth hitpoint

Region IV Polar angles between 70 and 140

representing the area covered by all four barrellayers

The hit information in Region I contributes to theforward tracking systems which basically use astraight line fit before and after the dipole There-fore one or two hit points from the MVD is al-ready sufficient The design goal of at least fourhit points has been applied to the three remainingregions They are connected to the central trackingpart for which a helix fit must be performed Themost important part for many physics channels isgiven by Region II which also includes the kine-matic region of charged decay modes of the groundstate D mesons Compromises had to be done inRegion III in order to obtain a feasible solutionfor the detector assembly However the slightlyreduced performance can be counter-balanced bythe outer tracking system The crossing target pipeaffects the main barrel part defined by Region IVand results in an acceptance cut A full coverage of100 is reached for the disk layers (Region II ) thecrossover between the barrel and the forward part(Region III ) and down to a minimum polar angleof θ = 45 in the forward direction (Region I )

Figure 620 illustrates the percentage coverage withdifferent hit counts in the four detector regions InRegion I a hit count of one and two dominates(565) Contributions from the inner rims of thefour pixel disks start at 3 45 65 and 8 respec-tively A maximum number of 14 hit points as wellas the highest averaged value are obtained in RegionII In this area four and five hits contribute with onethird each while the proportion of lower hit countsis only slightly above 10 Within the covered solidangle of Region III mostly three hit points (41)are obtained However at least four hit counts arestill reached in roughly half of the area The cover-age in Region IV is clearly dominated by four hitpoints with a contribution of 50 While the per-centage of radial sensor overlaps which deliver alarger hit count is in the order of 15 the areaaround the target pipe with a reduced number ofhit points covers roughly 29 of the full acceptancein this region


Particle ChargeRelative coverage within 3 lt θ lt 150 []

momentum conjugation Number of MVD hit points

1 2 3 4 5 6 Total

015 GeVc+ 63 76 223 385 169 36 950

minus 45 90 210 384 171 34 932

05 GeVc+ 61 64 230 397 169 38 949

minus 54 71 224 409 166 36 945

10 GeVc+ 60 64 227 400 171 37 952

minus 56 69 224 408 166 38 950

15 GeVc+ 60 64 226 402 171 38 952

minus 57 68 224 408 166 39 952

Table 62 Relative detector coverage in the active detector region obtained with pions of both charge conjuga-tions

Figure 621 Distances of the first hit point to theorigin (000) obtained for slow pions (p = 150 MeVc)

Besides a sufficient number of MVD hits for eachparticle track a minimum distance of the first hitpoint with respect to the nominal IP is crucial forthe physics performance of the detector A 2D illus-tration of the absolute distance of the first hit pointto the origin is given in figure 621 on the left Nextto it on the right corresponding 1D profiles alongthe the polar angle are added They show the min-imum and the averaged value obtained within thefull azimuthal angle Results indicate that a dis-tance of less than 4 cm can be reached within awide range of the covered solid angle Minimumdistances to the nominal interaction point are ob-tained in the barrel part and at polar angles around30 are in the order of 2 cm

663 Material Budget

The introduction of the MVD is inevitably con-nected to some detrimental effects given by scat-tering and energy straggling of charged particles aswell as the conversion and attenuation of photonsinside the detector material The impact of the in-troduced material can be quantitatively describedby a resulting fractional radiation length XX0which is accumulated in all traversed volumes j

XX0 =sumj

ρj middot LjX0j

(614)

where X0j and ρj are the specific radiation lengthand the density of the material defined for the vol-ume j respectively and Lj corresponds to the tra-versed path length therein In the further discus-sion the corresponding value of the material budgetwill be given in a percentage of one full radiationlength X0 if XX0 lt 1 or in multiples of X0 ifXX0 ge 1

The following studies about the material budget arebased on the comprehensive MVD model which hasbeen introduced in chapter 62 For the extrac-tion of a material map virtual particles (ldquogeanti-nosrdquo) were propagated through the detector Thefictitious ldquogeantinordquoparticle undergoes no physicalinteractions but flags boundary crossings along itsstraight trajectory Two million events were sim-ulated starting from the origin (xyz) = ](000)with an isotropic emission over the polar angle θand the azimuthal angle φ Further on the effec-tive path lengths in all traversed volumes (Lj) wereidentified for each propagated geantino The frac-tional radiation length (XX0) was then calculated


Figure 622 2D map of the material budget (a) and extracted 1D profiles along the polar angle (b)

based on the given material definition of the vol-umes according to equation (614)

In figure 622 (a) the obtained material map of theMVD is given in the representation of the frac-tional radiation length which is plotted colour-coded against both initial emission angles Corre-sponding profiles along the polar angle are shown infigure 622 (b) They are obtained for a bin size of∆θ = 1 Resulting numbers for the average and themaximum fractional radiation length within eachinterval are shown at the top The standard devi-ation of all values normalised to the correspondingmean value is plotted below This ratio is a measureof the isotropy in the material distribution alongthe integrated azimuthal angle The obtained ma-terial distribution indicates a very low material bud-get of the MVD in the physically most importantregion at polar angles smaller than 140 Corre-sponding values stay mainly well below 10X0 andthe overall material map shows a rather isotropicoccupancy with relative fluctuations of not morethan 10 Larger deviations are located betweenθ = 30 and θ = 50 as well as in the very backwardregion (θ lt 150) which generally exhibits an in-creased material budget with peak values of severalradiation lengths Further details of the materialmap will be discussed in the following paragraphs

First of all the focus is given to local hot spotsie regions with an enhanced fractional radiationlength Figure 623 illustrates the azimuthally inte-grated distribution of such hot spots along the polarangle The given relative abundance is defined asthe fraction of counted tracks Nall for which theobtained value of XX0 stays in the specified range

Figure 623 Relative abundance NspotNall of hotspots in the material map along the polar angle Spec-ified threshold values are given inside the diagrams

divided by the overall number of tracks Nall to beexpected within an interval of ∆θ = 5 The lat-ter Nall is given by the percentage of the coveredarea in the φminus θ diagram multiplied with the totalnumber of simulated events

Different intervals for the fractional radiation lengthwere chosen to classify hot spots with respect tothe average value obtained in the sensitive detectorregion (θ lt 140) A lower limit for the definitionof a hot spot is given by

〈XX0〉+ 2 middot 〈σθi〉 = 91 (615)

where 〈XX0〉 and 〈σθi〉 represent the mean val-ues of the fractional radiation length and the stan-dard deviation within the ∆θ intervals respectivelyBoth of them are extracted from the 1D profilesshown figure 622 Finally four different categorieshave been specified The first category (I) defined in


an interval of 10 lt XX0 lt 25 rather describeslarger fluctuations to the averaged value Moderatehot spots of category II are in the order of 25X0

and 50X0 The last two categories III and IVaccount for hot spots with a very large value belowand above one full radiation length X0 respectively

Compiled results in figure 623 indicate that hotspots of category II to IV are basically outside thesensitive detector region of the MVD An excep-tional contribution of moderate hot spots (cate-gory II) in the order of 5 can be located betweenθ = 35 and θ = 40 All other hot spots of the sec-ond third and fourth category are located in theupstream region between θ = 140 and θ = 170Fluctuations of the material map included in cate-gory I are mainly spread over the sensitive detectorregion The maximum of this distribution shows upat θ = 145

In order to obtain a better understanding of theresulting pattern in the material map it is usefulto perform a separate analysis of different detectorparts Therefore only the individual componentsof interest were switched on during the simulationIn a first step the four main substructures of theMVD model have been chosen These main partsare (cf figure 63)

SiliconSensor elements and associated readout chips

SupportGlobal and local support structures

RoutingService structures for sensors and front-endchips including supply and data cables as wellas cooling pipes

ComponentsElectronic and cooling connectors as well as ad-ditional SMD components

The individual contributions of these substructuresare disentangled in figure 624 In comparison withfigure 622 it is evident that the most significantimpact on the material map originates from therouting part It delivers the obvious jump to alarger fractional radiation length at polar anglesabove 140 and also includes hot spots with max-imum values of several X0 These enhancementsare related to the fixed routing scheme of the innerbarrel layers along the opening beam pipe in up-stream direction in combination with the generallyincreased incident angle of tracks traversing the cir-cularly arranged service structures In the forwardregion (25 lt θ lt50) the routing of the pixel disk

services at the top and bottom is clearly visible Itresults in a partial increase above 10X0 (cf fig-ure 623) thus defining the maximum material bud-get in the sensitive detector region Besides averagevalues therein level around 5X0

Compared to the routing part maximum values ofall other sub-structures in the 2D distribution fallshort by more than one order of magnitude Sup-port structures contribute with 1X0 to 3X0 inthe active detector region Moreover higher val-ues reflect the load pick-up of the main buildingblocks which is concentrated in upstream directionat polar angles above 140 A maximum of ap-proximately 5X0 and 10X0 is given below (for-ward part) and above (barrel part) respectivelyThe most homogeneous material occupancy around2X0 is given by the silicon part It shows a typ-ical cos θ-modulation reflecting the incident angleof straight lines originating from the nominal inter-action point A further increase at smaller polarangles is related to the enlarged number of indi-vidual detector layers in forward direction Finallythe resulting pattern of the introduced componentsdefines a very anisotropic distribution While theaveraged number remains relatively small maxi-mum values of around 5X0 are reached at singularpoints Most of them are related to cooling connec-tors needed to change from steel pipes to flexibletubes

For the evaluation of different hardware develop-ments in terms of their impact on the overall ma-terial budget a slightly modified representation isused which re-orders the different volumes of theMVD model to functional groups

SensorsSilicon pixel and strip detectors

ElectronicsReadout chips SMD components and elec-tronic connectors

CablingRouted supply and data cables (core and insu-lation)

CoolingCoolant pipes and tubes as well as cooling con-nectors

SupportAll support structures (as defined before)

The stacked diagram in figure 625 (a) illustratesthe individual contributions for each of these func-tional detector parts along the polar angle Plot-


Figure 624 Maps of the contributing material budget in terms of the fractional radiation length (XX0) forthe four main parts of the model as illustrated in figure 63 Note the different colour scales for XX0

Figure 625 Contributions of different functionalparts of the MVD to the overall material budgetStacked 1D profile along the polar angle (a) and per-centage of the different parts within the sensitive detec-tor region (θ lt (140)) (b)

ted values represent the azimuthally averaged frac-tional radiation lengths XX0 within an intervalof ∆θ = 1 Besides some accumulation effects inthe very forward region the proportions between allparts remain relatively constant within the sensitivedetector region (θ le140) In figure 625 (b) theirpercentage on the overall budget is presented in apie chart The biggest amount of material is deliv-ered by the cables (38) and the support structures(29) The dominating part (70) of the materialbudget of the cabling part splits into 78 and 22for the supply and the data cables respectivelyThe introduced silicon sensors assuming standardtotal thicknesses of 200 microm and 300 microm for thepixel and the strip detectors respectively accountfor only 15 The cooling system adds another 14to the material map and the impact of the electron-ics stays in total rather small (5) However itshould be recalled that the discussion on integrateddistributions does not supersede a careful evalua-tion of the 2D map as done before

In summary the obtained composition of the mate-rial budget is typical for inner silicon tracking sys-tems of hadron physics or high-energy physics ex-periments In case of PANDA complications for aminimised material budget are given by the over-all routing scheme which requires the upstreamrouting of all services thus implying eg folded ser-vice lines for the pixel disks and the DAQ con-cept which is based on a triggerless readout thusincreasing the material due to the higher function-ality of the readout chip and the larger amount ofdata to be handled Even though the total amountof material for all service structures can be kept inan acceptable range This fact results from an op-


timised powering scheme for the readout electron-ics which decreases the required core cross sectionof the supply cables and the use of ultra-thin ca-bles based on an Aluminum core Both reduce theoverall material budget by roughly 45 Moreovera significant minimisation is given by the use oflightweight carbon support structures with a tai-lored design ie larger cut-outs and sandwichedstructures with Rohacell Further improvementswhich are not included in the presented studies arerelated to thinned silicon detectors the use of alu-minum supply cables surrounded by just a thin cop-per layer and the implementation of a higher dataconcentration The maximum achievable reductionin best case is expected to be in the order of 25A detailed discussion of these point can be foundin [35]

664 Rate Estimation

For the MVD as innermost subdetector of PANDAan evaluation of the expected count rates in differ-ent detector regions is from utter importance for thedevelopment of all readout electronics Extractednumbers result from the initial conditions of theexperiment ie the particle distribution in the re-action channel (cf figure 22 chapter 23) and thebeam-target interaction rate taken as basis for thecalculations In a first approximation the lattercan be correlated with the nominal PANDA inter-action rate which is given by 2middot107 sminus1 With thetime-averaged information it is possible to extractthe hit occupancy in different detector regions andthe overall data load expected to be handled by theMVD For a reasonable estimate on peak rates oc-curring on single readout channels and onfrontend level variations of the beam-target inter-action on a shorter time scale must be considered

A detailed count rate study for pp reactions hasbeen performed at minimum (15 GeVc) maximum(15 GeVc) and an intermediate beam momentumof 5 GeVc [35] A large statistical sample of twomillion DPM events were simulated for each of thethree setups Generated particles were propagatedthrough the detector within the PandaRoot frame-work Moreover the created hit points in the de-tector were digitised as described in chapter 632In the further analysis only binary hit informa-tion ie the IDs of sensor and front-end as wellas the number of the readout channel was storedthus allowing an unambiguous allocation of each hitpoint in the detector The summed hit counts werethen multiplied with the nominal PANDA interac-tion rate in order to obtain the time-averaged count

rate The given number for an individual readoutchip results from an integration of all associatedreadout channels In the same way average countrates for individual sensors are defined by the con-tribution of all connected readout chips The hitoccupancy for different detector parts is then givenby the accumulated count rates of all sensor ele-ments therein

Obtained results for the count rate distributionon front-end level and the individual contributionsof single readout channels are summarised in fig-ure 626 Due to the distinct emission of particlesin forward direction with increasing beam momen-tum the highest count rates inside the MVD occurat maximum beam momentum in the inner pixeldisk layers Upper limits of three million countsper second (Mcps) are obtained for individual pixelreadout chips located at the inner sectors of thesmall pixel disks Also at the inner rims of the sub-sequent pixel disks average count rates stay at ahigh level between 19 Mcps and 25 Mcps At outerradii values drop quickly by one order of magni-tude For the two innermost pixel disks the situa-tion is more challenging because all the front-endshave to cope with a high count rate of more than1 Mcps These special conditions require a sophis-ticated cooling concept for the readout electronics(cf chapter 561) which is crucial for a stable op-eration The highest count rates for readout chipsin the strip disks are in the order of 03 Mcps Theshorter strips at one sensor edge of the trapezoidalsensors lead to a lower count rate for the front-endat the outermost position This effect is mirroredat the opposite sensor side

In contrast to the forward part the occupancyin the barrel part increases with lower beam mo-mentum due to the increased number of particlesemitted at larger polar angles The distributionsat highest and lowest beam momentum are plot-ted in figure 626 The impact of the slow elas-tic recoil protons results in a local maximum forfront-ends located just behind z = 0 which corre-sponds to the observed peak in the overall parti-cle distribution shown in 22 Due to the orienta-tion along the beam axis the average count ratesof long strips stay for one sensor always above theones of short strips Upper limits for the strip partare obtained in the third barrel layer at minimumbeam momentum Assuming an interconnection ofthe two foremost sensors which is one of the possi-ble solutions envisaged to reduce the total numberof readout channels maximum values for the stripfront-ends stay slightly below 1 Mcps This valueis slightly exceeded for the foremost readout chips


Figure 626 Extracted average count rates for antiproton-proton reactions obtained with the DPM generatorThe distribution over all individual channels of the strip and the pixel front-end with the highest occupancy areshown in the frames on the left at top and bottom respectively Arrows indicate the position inside the MVDResults for all readout chips in the different detector layers are shown on the right (All results are based on anassumed interaction rate of 2 middot 107sminus1)

of the first pixel layer thus defining the maximumaverage count rates occurring in the overall barrelpart In the upstream part (z lt 0) obtained valuesdo not exceed a level of 01 Mcps

Pixel and strip readout chips with the maximumoccupancy also contain readout channels with thehighest individual count rates In case of the strippart the distribution over all connected readoutchannels ranges between 7000 counts per second(cps) and 8000 cps For the pixel part a more

anisotropic occupation on the pixel readout chipis obtained Values for single pixel cells differ byroughly one order of magnitude with an upper limitof roughly 1200 cps The increased count rate atthe top and the bottom of the pixel matrix stemsfrom an enlarged pixel size which is implementedto counter-balance the passive edge of the readoutchip

The most important results of the presented countrate study are compiled in table 63 For the inte-


Barrel Forward Sensor Front-end Readout

part part level level channel

Pixel part

Digitised hits 67 7rarr 42 166 7rarr 322 le 089 le 029 2 middot 10minus4

Ndig [106] (15 7rarr 15) GeVc (15 7rarr 15) GeVc 〈le 021 〉〈 004 〉〈 35 middot 10minus6〉

Average count rate 233 7rarr 364 le 89 le 29 0002

〈 Ndig〉 [Mcps] (15 7rarr 15) GeVc 〈le 21 〉〈 04 〉

Expected data rate 2200

45 17 minus〈Ndig〉middotfDAQ [MBs] 〈 12 〉〈 2 〉

Estimated peak rate minus minus gt 40 001

〈Ndig〉middotfpeak [Mcps] lt 145

Strip part

Digitised hits 211 7rarr 176 50 7rarr 84 030 009 8 middot 10minus4

Ndig [106] (15 7rarr 15) GeVc (15 7rarr 15) GeVc 〈 010 〉〈 002 〉〈 15 middot 10minus4〉

Average count rate 418 7rarr 416 48 15 0013

〈Ndig〉middotfro [Mcps] (15 7rarr 15) GeVc 〈le 16 〉〈le 03 〉


29 9 minus〈Ndig〉middotfro middotfDAQ [MBs] 〈 10 〉〈 2 〉

Estimated peak rate minus minus gt 20 gt 002

〈Ndig〉middotfro middotfpeak [Mcps] lt 55 lt 007

Table 63 Main results of the count rate study performed with 2 million DPM events Average count rates areobtained with the nominal interaction rate of 2middot107 sminus1 ie 〈Ndig〉 = 10 middotNdigmiddot sminus1 Given numbers at sensorfrontend and channel level represent the maximum values obtained for a single element Mean values of allelements in the corresponding setup are indicated with 〈〉

grated pixel and strip part an upper limit on the av-erage count rate is given by 360 Mcps and 420 Mcpsrespectively In case of the strip part final numberscontain a factor of fro = 16 [35] which consid-ers an induced charge sharing between neighboringchannels ie a minimum number of two activatedstrips per hit Such technique based on a smallersubstructure with interstitial floating strips is en-visaged for the silicon strip detectors in order toimprove the spatial resolution Taking into accounta size of 40 bit for the output format of the hit data(cf chapter 541) it is possible to estimate thecorresponding data load By introducing a scalingfactor fDAQ = 6 byte 5 obtained numbers for the av-erage count rate can be translated into a data rategiven in megabytes per second (MBs) The maxi-mum data load of roughly 47 GBs is expected athighest beam momentum at which the contribu-tion of the pixel and the strip part are in the sameorder of of magnitude

Fluctuations of the total number of registered hitson a short timescale deliver increased peak rateswhich have to be buffered by the readout electron-ics As a first approach the intrinsic change of theinstantaneous luminosity profile during the opera-tion cycle can be taken to deduce a lower limit onthe occurring peak rates According to model cal-culations for the HESR [36] it is given by an increaseof fpeak asymp 138 which results from the ratio of theluminosity at the start of the data taking period tothe cycle-averaged value Additional fluctuationsmay be introduced by variations of the effective tar-get thickness and the microscopic structure of thebeam An initial rudimentary study for the use of apellet target at PANDA [37] allows for a very roughestimate of an upper limit which results in a scalingfactor of fpeak asymp 5

5 Compared to the 5 bit format defined in chapter 541an additional safety bit was added to estimate an upper limit


Pixel Strip Barrel Strip Disks

Readout size 100times100 microm2 130 microm 65 microm

Stereo Angle ndash 90 30

Noise σn 200 eminus 1000 eminus 1000 eminus

Threshold 1000 eminus 5000 eminus 5000 eminus

Charge Cloud σc 8 microm 8 microm 8 microm

Table 64 Simulation parameters overview

Further simulations included in [38] [39] havebeen performed to study the possible effects ofantiproton-nucleon reactions on the expected countrates inside the MVD Results indicate that the ex-tracted hardware specifications based on the num-bers as given in table 63 suffice for a complianceof the required detector performance under theseconditions

67 Resolution and PerformanceStudies

This section contains studies dealing with the per-formance of the MVD as standalone detector and inthe whole PANDA setup For comparable results aset of default parameters see table 64 has beenused

671 Hit Resolution

A good position resolution of each hit on a sensoris mandatory for a good momentum and vertex re-construction To have a clean sample only chargedpions with a momentum of 1 GeVc have been usedcovering the whole azimutal angle range while thepolar angle covers 7 to 140 The figure 627 showsthe resolutions per sensor type and by cluster mul-tiplicity For the case of one digi per cluster theresolution is limited by the readout structure size(100 microm for the pixels 130 microm and 65 microm for thestrips) When the charge weighted mean is usedthe achieved single hit resolution largely improves(σx = 69 microm for the pixels and σx = 124 microm forthe strips)

672 Vertexing Performance on Pions

Several scans were performed to characterize thevertexing performance of the MVD Four pions(two π+ and two πminus) of 1 GeVc were propagatedthrough the detector from a common vertex dis-tributing homogeneously the four particles along

the θ and φ ranges (θ isin [10 140] φ isin [0 360])The vertices were set in different regions to mapthe resolution and to perform systematic studies(10000 simulated events per point)

bull longitudinal scan moving the vertex along thelongitudinal (z) axis in the range [-1+1] cm

bull circular scan positioning vertices along a circlewith a radius 01 cm in the transverse (x-y)plane

bull radial scan changing the distance between(000) and the vertex moving along a radiusin the transverse plane

These studies were realized using the MVD theSTT the GEMs and the forward tracker Thetracking was performed merging the informationprovided by the barrel and the forward spectrom-eters A Kalman filter was applied to tracks mea-sured in the barrel spectrometer correcting for en-ergy loss and scattering occurring in the detectors(see section 64) Forward tracks were reconstructedusing the ideal pattern recognition and smearing thetrack parameters according to resolution and effi-ciency foreseen for the forward tracker Finally thePOCA vertex finder (see section 644) was appliedto track candidates obtaining the reconstructed po-sition of the vertices

All the scans showed a similar behavior the vertexresolution achieved for the x coordinate is worsethan y This is not expected due to the φ sym-metry of the MVD geometry The reason of thisdiscrepancy is the design of the inner barrel layersof the MVD which is limited by mechanical con-straints The top and bottom regions close to thenominal interaction point cannot be covered withpixel modules in the range θ isin [50 90]

This results in a different precision in the determi-nation of the coordinates of the reconstructed ver-tices which is significant in this portion of solidangle Figure 628 shows the x and y resolutionsobtained with selected θ values for the four pions


diffnxoneEntries 94388

Mean 109e-05

RMS 0002158

[cm]reco-xmcx-001 -0005 0 0005 0010

200400600800

100012001400


Mean 109e-05

RMS 0002158 =1clustPixel Resolution n diffnyoneEntries 82649

Mean 1127e-05

RMS 0002886

[cm]reco-xmcx-001 -0005 0 0005 0010

200

400

600

800

1000

diffnyoneEntries 82649

Mean 1127e-05

RMS 0002886 =1clust

Strip Resolution n

diffnxmulEntries 51326

Mean -3275e-06

RMS 0001379

ndf 2

χ 4465 18

Constant 165plusmn 1970

Mean 5369e-06plusmn -9899e-06

Sigma 00000074plusmn 00006904

[cm]reco-xmcx-001 -0005 0 0005 0010

500

1000

1500

2000

2500diffnxmul

Entries 51326

Mean -3275e-06

RMS 0001379

ndf 2

χ 4465 18



Sigma 00000074plusmn 00006904 gt1clustPixel Resolution n

mmicro 007)plusmn = (690 σ

diffnymul

[cm]reco-xmcx-001 -0005 0 0005 0010

500

1000

1500

2000

2500

diffnymul gt1clust

Strip Resolution n

mμ 01)plusmn = (124 σ

Figure 627 Resolution on pixel and strip sensors for 1 GeVc pions

Figure 628 Resolutions obtained selecting a fixedvalue for the θ angle and using an uniform distributionfor the φ angle

A detailed analysis of the origin of this different be-havior can be found in the appendix C The studiespresented this section were performed covering ho-mogeneously a wide θ range This is the reasonwhy in the results a significant difference appearsbetween the x and the y vertexing performances

Most of the physics channels of interest for the ex-periment foresee mainly decays in the forward re-gion of the detector therefore the x-y different per-formance at big θ angles will not influence most ofthe physics studies For example all the results ofthe next section (68) show the same resolution forthe x and y vertex reconstruction

Results of the Scans

The longitudinal scan (figure 629) shows that ver-tex resolution values are quite stable in the range[-5+5] mm where the pattern recognition for pri-mary particles is foreseen to work efficiently Thecircular scan (see figure 630) confirms a good φ-uniformity of the performance expected on the ba-sis of the detector geometry In figure 631 the ef-fects of a movement along a radius with φ = 0

and z = 0 cm are shown The resolutions are notseverely influenced by this kind of shift as long asone stays at distances below 1 cm from the nomi-nal interaction point There the vertexing starts toshow worse performances

The momentum resolution is much less effected bychanges in the position of the vertex and it staysconstant within a few per mill in all the performedscans (see figure 632)


Figure 629 Vertexing results obtained placing de-fined vertices in different points along the longitudinalaxis of the MVD

Figure 630 Scan of the vertexing behavior movingvertices on a circle of radius 1 mm laying in the trans-verse plane

Figure 631 Mapping of the vertexing results chang-ing the distance from the nominal interaction pointmoving along the radius corresponding to φ=0 andz=0 cm

Figure 632 Momentum resolution values obtainedfor π+ and πminus with the three scans

68 Physics Channels Analysis

The tasks of the MVD are mainly to provide hightracking performance close to the interaction zoneWe report three physics benchmark cases where theperformance of the MVD is monitored First wehave particles decaying directly in the interactionpoint (Ψ(2S)rarr Jψπ+πminus) and in the second caseparticles which decay further away from the inter-action point (D mesons)

681 Detector Setup

In the physics channels we chose the following geo-metric setup Active components


bull MVD 4 barrel layers and 6 disks

bull STT 15 m long with skewed layers

bull GEM 3 stations

bull FTS 6 forward tracking straw stations

For these elements hit digitization and reconstruc-tion is performed In the barrel spectrometer partthe combined STT+MVD +GEM pattern recog-nition with a Kalman Filter track fitting is usedwith the default particle hypothesis being muonsIn the forward spectrometer a Monte-Carlo truthbased tracking is applied with a 95 efficiency andgaussian smearing of σpp = 3 in momentum aswell as σv = 200 microm in position Realistic parti-cle identification algorithms are not used in orderto have results independent from them To identifyparticles the Monte-Carlo information is used

The following components were present as passivematerial thus contributing to the material beingpresent excluding the time consuming digitizationprocesses of these components not being relevant inthese studies

bull Pipe Beam-Target cross (passive)

bull Disc DIRC

bull Barrel DIRC

bull EMC crystals

bull Solenoid and forward dipole magnet includingflux return yokes and muon chambers

The simulations of physics channels were donewith the PandaRoot ldquotrunkrdquo revision number 12727from July 18 2011 using the ldquomay11rdquo release of theexternal packages

682 Benchmark Channelpprarr ψ(2S)rarr Jψπ+πminus

Jψ rarr micro+microminus

The first test has been performed reconstructing thefollowing reaction

pprarr ψ(2S)rarr Jψπ+πminus rarr micro+microminusπ+πminus

Running a full analysis and performing a vertexfit on the Jψ candidates it was possible to de-termine the vertex resolution Figure 633 showsthe distributions of the x and y coordinates ofthe reconstructed vertices The resolution in z isslightly worse (σz = 64 microm while σx = 43 microm and

σy = 42 microm) due to the forward boost of the decayThe overall vertex reconstruction is excellent

Figure 634 shows the mass distribution obtainedfor the Jψ and the ψ(2S) candidates respectivelyThe masses obtained fitting these distributions arein full agreement with the expected values

Figure 635 shows the missing mass of the pp rarrπ+πminus reaction at the ψ(2S) peak from which theJψ signal emerges The Jψ signal obtained inthis way is much narrower than that obtained bythe invariant mass of the lepton pairs (figure 634)

Jψ rarr e+eminus

The decay of the Jψ into e+ eminus has been stud-ied in the same situation This decay is more ef-fected by multiple scattering than the muonic oneTherefore some selections on the reconstructed can-didates can improve the final results Only π ande candidates matching the allowed phase space re-gions6 were taken into account for the rest of theanalysis Figure 636 and figure 637 show the MCprediction for the polar angle and momentum distri-butions compared with what is obtained from thereconstruction Two colored bands highlight thespecific cuts applied on the kinematics of the trackcandidates Applying these selections Jψ candi-dates have been reconstructed The distributionsof the reconstructed vertex positions are shown infigure 638 Figure 639 shows the invariant mass ofpossible Jψ candidates and the missing mass builtby the measured pions (which obviously does notdepend on the the Jψ decay mode)

Figure 640 shows the distribution of the recon-structed ψ(2S) vertex positions These vertices arecalculated combining four particles Therefore moreconstraints are applied than in the case of the Jψvertex This results in a better z resolution due tothe forward boost of the system These considera-tions are based on the fact that both the ψ(2S) andthe Jψ decay immediately in the interaction ver-tex The invariant mass of the ψ(2S) candidates areshown in figure 641 where the results of a Gaussianfit on the peak are reported Since ψ(2S) candidatesare obtained combining Jψ ones with a π+ anda πminus which both have to satisfy track quality re-quirements the final total number of reconstructedψ(2S) is smaller than the Jψ available statistics

6 This cut was based on the MC-true phase space distri-butions for this decay


Figure 633 Distributions of the Jψ vertices reconstructed from their decay in the nominal interaction point

Figure 634 Mass distribution for Jψ and ψ(2S) candidates

Figure 635 Distribution of the Jψ mass obtained with the missing mass method knowing the initial ψ(2S)state


Figure 636 Pion polar angle and momentum distributions from MC-truth (left) and from the reconstruction(right) In the second plot the kinematic cut is highlighted by a colored band

Figure 637 Electron-positron polar angle and momentum distributions from MC-truth (left) and from thereconstruction (right) In the second figure the reconstructed distribution is superposed to the kinematic cutapplied during the analysis

Figure 638 Reconstructed Jψ vertices


Figure 639 Jψ Reconstructed mass (left) and missing mass (right) distributions

Figure 640 Coordinates of the reconstructed ψ(2S) vertices

Figure 641 Mass distribution for the ψ(2S) candidates


Figure 642 D+Dminus maximum opening angle de-pending on the p beam momentum

683 Benchmark Channel D mesons

In order to explore charm physics in PANDA thereconstruction of D mesons is a crucial point Thebiggest challenge is to reduce the background fromthe signal channels which have typically a low crosssection Because the decay lengths are cτ = 312 micromfor the charged and cτ = 123 microm for the neutralground state D mesons a selection by the decayvertex is envisaged

6831 pprarr D+Dminus

The charged ground state D mesons decay by 94branching fraction into a final state with threecharged particles (Kππ see [40]) It is convenientto have only charged particles in the final state tobe less dependent on other detector components inthis study In the exit channel six charged tracksthen have to be reconstructed

pprarr D+Dminus rarr Kminusπ+π+K+πminusπminus

For this study 99200 signal events have been sim-ulated at 9 GeVc incident antiproton momen-tum enabling a large maximal opening angle be-tween the two D mesons (see figure 642) Fur-thermore a set of 344525 non-resonant backgroundevents was simulated featuring the same final stateK+Kminus2π+2πminus

After the reconstruction 609 of the original Dmesons survived with 338 of additional combi-natorial background The exact numbers are com-piled in table 65 A rather wide mass window cutof plusmn 150 MeVc2 around the nominal charged Dmass (taken from [40]) was applied to reduce the

m]microct [

0 500 1000 1500 2000

1

10

210

310

mmicro19 plusmn = 3122τc

Figure 643 D+ and Dminus decay lengths (PDG valuecτ = 3118 microm) [40] obtained directly with the POCAfinder without any correction

non-resonant background by a large factor Cut-ting on the vertex z coordinate (POCA) of the Dcandidates (478 microm lt z) and on the ldquodistancerdquo(see section 644) calculated by the POCA finder(d lt 300 microm) much of the combinatorics as wellas the non-resonant background is removed Thevertex resolution values are given as the gaussianfitrsquos width in table 66 The three available ver-tex finderfitter algorithms have been used With aprecise vertex position it is possible to measure thedecay length of the Drsquos We found cτ = (3122 plusmn19) microm which is in agreement with the PDG valueof cτ = 3118 microm ([40] see figure 643) This in-cludes also a correction for the tracking acceptanceThe POCA finder gives a good resolution henceit was used to finally determine the mass resolu-tion (shown in figure 644) which is 204 MeVc2 byreconstruction only and 87 MeVc2 after the fitsThe combinations of both D mesons in each eventare fitted with a four constraint kinematic fit Herethe knowledge of the initial states four-momentumie the ideal beam is introduced After a cut onthe fitrsquos χ2 probability (p(χ2ndf) gt 0001) mostof the background sources are suppressed and theD mass resolution is improved significantly (see fig-ure 644)

An important number is the signal-to-backgroundratio The number of simulated events has to bescaled by the channel cross section accordinglyThe non-resonant background production cross sec-tion can be estimated to σNR asymp 234 microb (value at88 GeVc in [41]) and for the signal to 004 microb [42]Using the suppression factors from table 65 and the


D+Dminus comb nonres D0D0

comb nonres

Simulated events 99200 ndash 344525 98800 ndash 198125

Simulated Drsquos 198400 ndash (689050) 197600 ndash (396250)

Drsquos After reconstruction 120835 67039 688188 135024 27161 322771

plusmn150 MeVc2 Mass window 61579 21963 67645 102763 2232 23301

Vertex r and z cut 15004 689 1060 17255 194 543

4C kinematic fit 1058 25 0 1592 18 2

Suppression factor (fits amp cuts) 88middot10minus3 37middot10minus4 lt 15middot10minus6 12middot10minus2 66middot10minus4 62middot10minus6

Suppression factor (total) 53middot10minus3 13middot10minus4 lt 15middot10minus6 86middot10minus3 91middot10minus5 51middot10minus6

Table 65 D meson counts from different sources Signal channel combinatorial background (comb) andnon-resonant background (nonres) All cuts and fits are applied sequentially

D+ Poca PRG KinVtx

σxmicrom 569 861 469

σymicrom 563 848 461

σzmicrom 113 125 932

Dminus Poca PRG KinVtx




Table 66 D+ and Dminus vertex resolution values ob-tained with the available vertex fitters

]2M[GeVc17 18 19 2 21

210

310

Reco

Vtx-Cut

Vtx-Cut amp 4C-Fit

2=235 MeVcRecσ2=204 MeVcVtxσ

2 =87 MeVc4Cσ

Figure 644 D+ and Dminus invariant mass resolutionsafter vertex selection and after a fit constrained by thebeam four-momentum

final states branching fraction one can estimate

SB =fS middot σS middotBR

fBG middot σBG + fC middot σS middotBRasymp 056 (616)

D0 Poca PRG KinVtx



σz microm 884 949 902

D0

Poca PRG KinVtx



σz microm 884 941 893

Table 67 D0 and D0

vertex resolution values ob-tained with the available vertex fitters

6832 pprarr D0D0

Neutral ground state D mesons decay by 389 intoa two charged particle state (see [40])

pprarr D0D0 rarr Kminusπ+K+πminus

Having only two particles to form the D a higherreconstruction efficiency is expected due to thesmaller number of particles which have to be in-side the detectors acceptance Out of the 98900simulated signal events 683 where reconstructedwhile additional 137 of combinatoric backgroundis found The numbers of surviving events after thecuts are compiled in table 65 while the vertex res-olutions with the three fittersfinders are presentedin table 67 Similar to the charged D mesons onethe mass resolution achieved is 245 MeVc2 beforeand 58 MeVc2 after the fits (see figure 646) Ex-tracting the decay length of the neutral Drsquos we findcτ = (1197plusmn08) microm which agrees with the PDGrsquosinput value cτ = 1229 microm [40]

Calculating the signal-to-background ratio we ap-proximate the cross section for the signals by us-ing the charged D mesons case σ

D0D0 asymp 0004 microb


m]microct [

0 200 400 600 800 1000 1200 1400

1

10

210

310


Figure 645 Decay lengths of D0 and D0

(PDG valuecτ =1229 microm) [40]

]2M[GeVc17 18 19 2 21

210

310

Reco

Vtx-Cut

Vtx amp 4C-Fit


2 =58 MeVc4Cσ

Figure 646 D0 and D0

Mass resolution values afterVertex selection and after a fit constrained by the beamfour-momentum

and σBG asymp 85 microb (value at 88 GeVc in [41] seealso figure 647) for the non-resonant backgroundThe resulting signal-to-background ratio using thenumbers from table 65 is

SB asymp 003 (617)

684 Considerations About theVertexing

The physics channels studied in the previous sec-tions proof the MVD vertexing capabilities Theseresults are not affected by the different x-y resolu-tions mentioned in section 672 since all the de-cays are boosted forward where the vertexing per-

62 Definition of the DD Benchmark Channels

622 Hadronic Background Reactions

Compared to the total pp cross section of 60 mb the DD production cross section willbe suppressed at least by ten orders of magnitude as explained in Section 621 Forthis analysis the event pattern in the detector will be six charged tracks Thereforebackground reactions leaving a similar signature can fake the physics signal

Figure 62 gives an overview of channels which are considered to be major back-ground sources The cross sections have been derived from data samples mainly from

[GeVc] lab

p0 2 4 6 8 10 12 14 16 18

[m

b]

-210

-110

1

10

210

310

p totalp

6 prong

-3+3

0-3+3

-2+2K2

X cross sections p p Figure 62 Overview about possible background reactions and their total cross

section compared to the total pp cross section

bubble chamber experiments The six-prong events are those which have six chargedtracks in the final state ([121][122][123][124][125][126]) The next two channels con-taining only pions either with one additional pion 3π+3πminusπ0 or exactly six chargedpions 3π+3πminus ([122][123][127][128]) For the later the cross section is one order ofmagnitude smaller These channels have a rather high cross section and therefore agood PID within PANDA is necessary to suppress the pions sufficiently compared tokaons to detect the D mesons via their decay into charged kaons and pions

The channel K+Kminusπ+π+πminusπminus may be the largest background source This chan-nel is still six to seven orders of magnitudes above the DD cross section estimatesfrom Section 621 but the experimental signature in the detector and the kinemat-ics are similar Only a precise tracking system which allows to clearly identify thesecondary vertices from the delayed D meson decays will allow to reduce this channel

115

Figure 647 Overview about possible 6-prong back-ground reactions and their total cross sections comparedto the total pp cross section [43]

formance of the MVD is homogeneous (see figurefigure 628) The reconstruction of Jψ candidatesshowed the vertex resolutions achievable with two-body leptonic decays (σx sim σy asymp 43 microm σz asymp60 microm) The study of the full ψ(2S) rarr Jψπ+πminus

reaction demonstrates the improvement of the z ver-tex reconstruction with a bigger number of con-straint (more tracks taken into account for the ver-tex determination) This effect is more evident inz the in the other coordinates due to the forwardboost of the system The detailed study of the ef-fect of several analysis techniques on the analysisof charged and neutral D mesons highlights howcrucial the MVD is when dealing with small crosssections and in the cases where the topology of thebackground are quite similar to the one of the sig-nal under study Vertex cuts result in being crucialfor the background suppression (for example in thecase of charged D mesons nearly a factor one thou-sand) since they can provide an efficient particleflagging on the basis of the decay lengths Whatwas shown is demonstrating the key role of vertex-ing Nonetheless more refined analysis strategieswill be implemented in the whole scope of the fullPANDA experiment


References

[1] S Spataro Simulation and event reconstruc-tion inside the pandaroot framework Journalof Physics Conference Series 119(3)0320352008

[2] Geant3 manual CERN program libraryW5013 1993

[3] Geant4 An object oriented toolkit for simula-tion in HEP Geneva httpgeant4cernch last visited Nov 22 2011

[4] VMC documentation 2010 httproot

cernchdrupalcontentvmc last visitedNov 22 2011

[5] I Hrivnacova et al The virtual Monte Carlo2003

[6] Rene Brun and Fons Rademakers Root - anobject oriented data analysis framework NuclInstr Meth A38981ndash86 1996 ProceedingsAIHENPrsquo96 Workshop Lausanne Sep 1996

[7] Torbjorn Sjostrand Stephen Mrenna and Pe-ter Z Skands A brief introduction to PYTHIA81 Comput Phys Commun 178852ndash8672008

[8] EvtGen documentation httpwwwslac

stanfordedu~langeEvtGen last visitedNov 22 2011


[10] SA Bass et al Microscopic Models for Ultra-relativlstic Heavy Ion Colllslons Prog PartNucl Phys 41255ndash369 1998


[12] C Hoppner S Neubert B Ketzer and S PaulA novel generic framework for track fitting incomplex detector systems Nucl Instr MethA620(2-3)518 ndash 525 2010

[13] RE Kalman A new approach to linear filter-ing and prediction problems J Basic Eng8234 1961

[14] R Fruhwirth Application of Kalman filteringto track and vertex fitting Nucl Instr MethA262(2-3)444 ndash 450 1987

[15] V Innocente et al The GEANE programCERN Program Library 1991 W5013-E

[16] A Fontana P Genova L Lavezzi and A Ro-tondi Track following in dense media and in-homogeneous magnetic fields PANDA reportPV01-07 2007

[17] L Lavezzi The fit of nuclear tracks in highprecision spectroscopy experiments PhD the-sis Universita degli Studi di Pavia 2007

[18] S Bianco Th Wurschig T Stockmanns andK-Th Brinkmann The CAD model of thePANDA Micro-Vertex-Detector in physics sim-ulations Nucl Instr Meth A654630ndash6332011

[19] CadConverter documentation Webpagehttppanda-wikigsidecgi-binview

ComputingCadConverter last visited June7 2011

[20] ISO 10303-11994 Industrial Automation Sys-tems and Integration Product Data Repre-sentation and Exchange - Overview and Fun-damental Principles International StandardISOTC184SC4 1994

[21] Th Wurschig Basic Conventions for theCAD-Modelling of the Micro-Vertex-DetectorPanda MVD-note 003 2008

[22] RA Boie V Radeka Centroid finding methodfor position-sensitive detectors IEEE TransNucl Sci 178543ndash554 1980

[23] R Fruhwirth A Strandlie and W Wal-tenberger Helix fitting by an extended rie-mann fit Nuclear Instruments and Meth-ods in Physics Research Section A Accelera-tors Spectrometers Detectors and AssociatedEquipment 490(1-2)366ndash378 2002

[24] PANDA STT Group Technical Design Reportof the PANDA STT 2011

[25] Pierre Billoir and S Qian Fast vertex fittingwith a local parametrization of tracks NuclInstr Meth A311139ndash150 1992

[26] P Avery Applied fitting theory IV Formulasfor track fitting CBX 92ndash45

[27] P Avery Applied fitting theory V Track fit-ting using the Kalman filter CBX 92ndash39

[28] A Giammanco Particle identification with en-ergy loss in the CMS silicon strip tracker CMSNote 005 2008


[29] G Schepers et al Particle identification atPANDA Technical Report Report of the PIDTAG 2009

[30] G Lindstrom Radiation damage in silicon de-tectors Nucl Instr Meth A512(1-2)30ndash432003 Proceedings of the 9th European Sym-posium on Semiconductor Detectors New De-velopments on Radiation Detectors

[31] A Vasilescu Fluence normalization based onthe niel scaling hypothesis In 3rd ROSE Work-shop on Radiation Hardening of Silicon Detec-tors DESY-PROCEEDINGS-1998-02 DESYHamburg Feb 12-14 1998

[32] A Lehrach Maximum luminosities withnuclear targets in HESR JahresberichtForschungszentrum Julich GmbH Institut furKernphysik 2006

[33] Marius C Mertens Der PANDA Mikro Ver-tex Detektor Entwicklung eines Labormesssys-tems Simulation der MVD-Betriebsparametersowie Untersuchungen zur Auflosung der Bre-ite des Dlowasts0(2317) PhD thesis Ruhr-Universitat Bochum 2010

[34] PANDA Computing Group Status of the Pan-daRoot simulation and analysis frameworkScientific report GSI 2007

[35] Th Wurschig Design Optimization of thePANDA Micro-Vertex-Detector for High Per-formance Spectroscopy in the Charm QuarkSector PhD thesis Universitat Bonn 2011

[36] F Hinterberger Beam-target Interaction andIntrabeam Scattering in the HESR TechnicalReport FZJ 4206 2006 ISSN 0944-2952

[37] A Smirnov A Sidorin and D KrestnikovEffective Luminosity Simulation for PANDAexperiment at FAIR Proc of COOL 2009Lanzhou THPMCP(002)1ndash3 2009

[38] L Zotti D Calvo and R Kliemt Rate studyin the pixel part of the MVD PANDA MVD-note 007 2011

[39] M Mertens Th Wurschig and R Jakel Countrate studies for the Micro-Vertex-DetectorPANDA MVD-note 004 2010


[41] J C Hill et al Strange particle productionin anti-p p annihilations at 88-GeVc NuclPhys B227387 1983

[42] AB Kaidalov and PE Volkovitsky Binaryreactions in anti-p p collisions at intermediateenergies ZPhys C63517ndash524 1994

[43] Rene Jakel Resolution Studies for the Mi-cro Vertex Detector of the PANDA Experimentand the Reconstruction of Charmed Mesons forSpecific Hadronic Channels PhD thesis TUDresden 2009


137

7 Project Management

The participating institutions in the PANDA MVDproject are located at the University of Bonn Ger-many (formerly at TU Dresden) at the IKP 1 ofthe Forschungszentrum Julich and the Universityand INFN of Torino As outlined in this Techni-cal Design Report these institutions share the loadof constructing an operational inner tracker systemfor the PANDA detector as discussed in the follow-ing In this chapter provisions taken or planned forquality control of components upon delivery fromthe vendors chosen and of subsystems will be dis-cussed

71 Quality Control andAssembly

Quality control of components upon delivery fromvarious vendors has already partly been discussedin the chapters dealing with the selection of certainparts for the MVD subsystems Routines for testprocedures such as radiation hardness and electri-cal uniformity and other performance parametershave been developed and are described in the pre-vious chapters They are briefly summarised in thefollowing

711 Pixel

The realisation of the hybrid pixel detector can bedivided into a sequence of elementary steps for dif-ferent elements sensors and readout chips mod-ules super modules completely equipped with cool-ing pipes and mechanics supports Additional in-frastructures are needed such as electrical-opticalconverters cables power supplies and off-detectorelectronics

Concerning sensors at present epitaxial raw waferscan be expected to be produced by ITME whichhas provided the material for the sensor RampD Theresistivity and the epitaxial layer thickness are di-rectly evaluated by the manufacturer to confirm thespecifications requested

FBK was the manufacturer of the pixel sensor pro-totypes produced in the RampD phase and performedthe quality control of each patterned wafer In par-ticular I-V and C-V curves are deemed very sensi-tive and good indicators of device quality Furthercontrol on test structures as C-V on MOS struc-

ture I-V trend on a gated diode and C-V inter-pixel measurements are foreseen to investigate spe-cific issues A visual inspection of metals and bond-ing pads of each sensor allows etching uniformityevaluation

Three sensor production batches (about 15wafersbatch) are sufficient to manufacture allpixel sensors needed for PANDA An additionalbatch can be foreseen for spare sensors Wafersamples from production processes will be used toobtain diodes for radiation damage tests to monitorthe radiation resistance uniformity using neutronsfrom the nuclear reactor at the LENA laboratoryin Pavia

After testing the sensor wafers are sent to manu-facturers such as VTT or IZM in order to obtain thepixel assemblies with the readout chips In partic-ular the Sn-Pb bump deposition is a delicate pro-cess but extensive screening was developed duringthe pixel detector realisation for the LHC experi-ments A failure during bump deposition will affectthe whole batch under work and most of these fail-ures can be detected with optical inspection whilecertain corrosion effects only show up after weeksA peculiar process that has to be applied to the sen-sor wafer for the PANDA MVD is the thinning inorder to remove most of the Cz substrate Failuresduring thinning can be destructive even breakingthe wafer An optical inspection allows to ascer-tain a positive result of the thinning process on themacroscopic level while I-V measurements will beused to exclude micro breaks Then sensor dicingis the last step before the coupling to the readoutchips Additional I-V tests assess the sensors toproceed to this working step Both VTT and IZMexhibit the capability to perform all the steps fortarget sensors as screened during the RampD phase

The readout chip wafers will be produced with adedicated engineering run They will be tested us-ing an additional probe card on a semi-automaticprobe station This equipment is already presentat INFN-Torino as well as in Julich The test isdelicate in particular due to the required degree ofcleanness to avoid contamination of the ASICs thatpass the test and proceed to the bump depositionprocess

The flip chip process joins pixel sensor and readoutchips to obtain assemblies (modules) Afterwardsa test is mandatory to have a fast feedback into the


bump bonding process even if this test endangersthe functionality of some of the finalised assembliesas this process has not yet been shown to producefully reliable constant yields The test has to beplanned at the vendorrsquos lab facilities because this isthe last step where the module can be reworked

In parallel with the assembly of modules produc-tion of all other components will be performed inorder to speedily proceed to super modules

Buses and cables are foreseen to be produced atCERN as a base solution Reliability tests bothwith visual inspection and electrical measurementsincluding accelerated aging processes are plannedusing suitable equipment at INFN-Torino

Mechanical supports carbon foam discs and barrelstaves will be assembled together with the coolingpipes at INFN-Torino using dedicated jigs Eachof these assemblies will be tested with a visual in-spection and verified with a precision measurementmachine

Dedicated jigs will be of fundamental importancefor the assembly of the super modules where thewire bonding process to connect readout chips tothe bus and the glueing between the layers are crit-ical steps at a stage where any mistake can poten-tially be dangerous to the whole system under workThis mounting is possible at the INFN-Torino ow-ing to the installations already used during the con-struction of the SDD of the ALICE experimentA final survey using a measurement machine ismandatory for mapping the system

The powering of the detector is based on a systemthat relies on several steps from power supplies toDC-DC converters (under study at CERN) to thevoltage regulators housed on the readout electron-ics

Off-detector electronics and DAQ are under devel-opment and based also on electrical-optical convert-ers under development at CERN (GBit project)Julich is involved in the adoption of this design forPANDA

The MVD mechanics has been designed and evalu-ated using FEM analysis and prototypes have beenmade for confirming the proposed concept Stresstests and behaviour of the mechanical setup as afunction of humidity and temperature variationswill be performed at INFN-Torino using suitableequipment such as a climatic chamber

The cooling plant for the full MVD system ispresently under study at INFN-Torino

712 Strips

In the course of prototyping for the strip sen-sors two manufacturers have demonstrated theirability to deliver double-sided sensors of suitablequality CiS (ErfurtGermany) was selected forthe first prototyping run of full-size barrel sen-sors whose performance during extensive testinghas been described in the strip part of this TDRFBK (TrentoItaly) also delivered sensors that havebeen used in the course of these tests

As detailed in the preceding text parameters suchas breakdown voltage leakage current etc will becontrolled past-production by the manufacturer inorder to guarantee the quality of the sensors deliv-ered according to the limits as discussed in section41 In the prototyping stage it was shown that themeasurement of global sensor I-V and C-V-curvesis already a good indicator for the quality of a sen-sor Hence it is planned to record these characteris-tics together with visual inspection data right afterthe delivery of the sensors Furthermore dose-leveldiodes from each wafer will be irradiated in orderto project the radiation fitness of the sensor fromthe respective wafer This can be done eg at thefission reactor of the Reactor Instituut Delft Tech-nical University DelftNetherlands

Assembly of the hybrid modules will be accompa-nied by tests after each assembly stage ie elec-trical test mechanicalstress test irradiation tests(on selected sub-samples) and thermal stress tests(ldquoburn-inrdquo)

713 Integration

The final integration of the components will be doneat the FAIR site The different pre-tested com-ponents will be shipped from the mounting sitesto Darmstadt Here a final electrical test will beperformed to ensure that the components were notdamaged during shipping After that the two half-shells of the MVD will be mounted starting fromthe outside to the inside with the strip barrel partfollowed by the pixel barrel part and the forwarddisks

After each mounting step the electrical and coolingconnections will be tested to ensure that none ofthe connections broke during the handling

Finally a full system test will be performed be-fore the detector can then be moved into PANDAand commissioned using the HESR beam interactingwith a PANDA target


72 Safety

Both design and construction of the MVD includingthe infrastructure for its operation will be done ac-cording to the safety requirements of FAIR and theEuropean and German safety rules Specific aspectsin the design are based on CERN guidelines thatare clearly appropriate for scientific installationsAssembly and installation ask for detailed proce-dures to avoid interference with and take into ac-count concomitant assembly and movement of otherdetectors Radiation damage aspects have to bechecked to allow the protection of all people in-volved in the operation and maintenance phasesA MVD risk analysis has to take into account me-chanical and electrical parts the cooling plant aswell as potentially hazardous materials and laserdiodes

The mechanical design of the support structures ofthe silicon devices has been checked by FEM anal-ysis and the material has been selected in con-cordance with the request of limited material bud-get Cables pipes and optical fibers are made ofnon-flammable halogen-free materials as well as arethe other components of the MVD In additionthe materials are chosen to be radiation tolerantat the radiation level expected in the PANDA en-vironment Here the specific CERN bibliographyconcerning the material classification in terms of ra-diation tolerance was used All supplies have appro-priate safety circuits and fuses against shorts andthe power channels have to be equipped with over-current and over-voltage control circuits DC-DCconverters have to be cooled to avoid overheatingand the power supply cables will be dimensionedcorrectly to prevent overheating Safety interlockson the electrical part have to be planned to preventaccidents induced by cooling leakage from other de-tectors Laser diodes are planned in the electrical tooptical converters hence safe housings and properinstruction (including specific warning signs) haveto be used The cooling system is a leakless cir-cuit based on water as cooling fluid working in atemperature range not hazardous to humans Aspecific control system including interlocks for elec-tronics manages the cooling circuit monitoring im-portant parameters such as temperature pressureetc When a malfunctioning is detected the systemacts to enable safety measures

73 Timeline and WorkPackages

The combined experience of the participating in-stitutions and the available personpower is suffi-cient to realise the PANDA MVD within the al-lotted timeline that is summarised schematically infigure 71 Note that the displayed time frame forthe project starts in 2010 Time spans needed forthe realisation of different detector components arelisted The sequence of steps within a block of tasksis usually interdependent so that subsequent stepscannot be parallelised While commissioning of thefull system is foreseen to proceed in the course ofthe first PANDA operation as a complete systemsmaller components will be available and functionalearlier for beam testing under realistic conditions inorder to verify the performance parameters

Risk items have been identified and fallback solu-tions are discussed in the chapters that deal withtechnically challenging components whose perfor-mance could not yet be fully predicted This in-cludes

bull the sensors of the hybrid pixel assembly wherethe EPI sensor research has been developed toa state where very thin sensors (asymp 100 microm) arewithin reach and can be expected to be pro-duced by selected foundries on a regular ba-sis and with high yield The process of bumpbonding these to the readout front-ends hassuccessfully been performed for several test as-semblies but cannot yet be deemed fully opera-tion for large-scale production A fallback solu-tion readily at hand is the replacement of thenovel sensors by more conventional solutionsfor pixelised sensors that have been used in thetest phase of the hybrid pixel sensor assemblyand rest on well-tested production proceduresat the cost of thicker Si layers

bull the free-running front-end for the strip part ofthe PANDA MVD Several solutions are underinvestigation and the need for a technical so-lution is common to all experiments plannedat FAIR Nevertheless an ASIC fulfilling thespecifications outlined in section 421 of thisreport has yet to be designed and produced insufficiently large quantities A PANDA-specificsolution would require an enhanced effort inASIC design which is not part of the presentproject planning and could presumably only beachieved with additional personpower and cost


Future milestones include the realisation of full-sizeoperational prototypes of the building blocks of thePANDA MVD from the components whose func-tionality has been discussed in the chapters of thisTDR and their assembly into operational small-scale systems These will then be merged to largergroups up to the final assembly of all parts for thePANDA experiment

The work packages of the project have been de-fined and distributed among the participating in-stitutions as outlined in figure 72 Where responsi-bilities are shared which is also the case among thegroups of Bonn and Iserlohn in the readout of thestrip sensors one of the partners in general providesthe work package management

The MVD group will continue to work in close col-laboration as demonstrated by regular meetings us-ing EVO and other electronic equipment as well asfrequent subgroup and specialised meetings at par-ticipating institutions Reports to the PANDA col-laboration will be provided on a regular basis asdone during the RampD phase

Acknowledgments

We acknowledge financial support from theBundesministerium fur Bildung und Forschung(BmBF) the Deutsche Forschungsgemeinschaft(DFG) the Gesellschaft fur SchwerionenforschungmbH GSI Darmstadt the ForschungszentrumJulich GmbH the Helmholtz-GemeinschaftDeutscher Forschungszentren (HGF) the IstitutoNazionale di Fisica Nucleare (INFN) and the Uni-versita degli Studi di Torino (Fisica) The supportof the National Funding agencies of all PANDAgroups and the European Union is gratefullyacknowledged

In addition this work has been partially supportedby the European Community Research Infrastruc-ture Integrating Activity HadronPhysics2 (grantagreement n 227431) under the Sixth and SeventhFramework Program DIRAC secondary beams thework-package WP26-ULISi and the FAIRnet I3

Additionally this work has been supported by theHelmholtz Association through funds provided tothe Virtual Institute ldquoSpin and Strong QCDrdquo (VH-VI-231)


Task NamePanda MVD

Pixel partSensorFront-End ASICController ChipModule Bus (Flex Hybrid)CablesHybridisationMounting

Strip partBarrel SensorsWedge SensorsFront-EndHybridController ChipCablesHybridisationMounting

Slow ControlOptical Data TransmissionClose To Detector ElectronicsCounting Room ElectronicsCables Power

Powering ConceptLV Power suppliesHV Power suppliesPower regulatorsCables

Mechanical StructuresGlobal SupportLocal Pixel Barrel Support CoolingLocal Pixel Disks Support CoolingLocal Strip Barrel Support CoolingLocal Strip Disks Support Cooling

CoolingDCS

MountingCommissioning

Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q12009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

Figure 71 Generalised timeline of the PANDA MVD project Time spans needed for the realisation of differentdetector components are listed The sequence of steps within a block of tasks is in general interdependent so thatsubsequent steps cannot be parallelised


Subdetector work packages Bonn Juumllich INFN-TorinoStrip partBarrel sectionSensors Electronics on detector Cables Mechanics and Assembly Cooling (strip barrel) Forward disksSensors Electronics on detector Cables Mechanics and Assembly Cooling (strip disks) Common tasks strip partMounting of modules Close to detector electronics El-Opt boards Pixel part Sensors Electronics on detector Assembly Bump bonding Mechanics and assembly Cooling (pixel side) Cables General tasksMechanics MVD assembly Cooling plant Power supplies Counting room electronics

Figure 72 Work packages the PANDA MVD project and their distribution among the participating institutions

143

Appendices

144

145

A The Bonn Tracking Station as a Validationfor the Simulation Framework

A1 The BonnTracking Station

The implementation of silicon devices such as mi-crostrips or pixels requires suitable tools to evalu-ate the main features either for sensors or for con-nected front-end electronics Among these studiesthose ones using a particle beam are very importantFor that purpose a beam telescope based on siliconstrips was designed and built by the Bonn MVDgroup It consists of four boxes positioned alongthe beam direction covering a maximum longitudi-nal range of nearly 200 cm Each box is providedwith silicon strips Single sided and double sidedsensor were used Both kinds have an active area of192 cmtimes 192 cm a thickness of 300 microm a pitch of50 microm and a 90 stereo angle Four scintillators aredisposed along the beam axis two of them beforeand two after the silicon sensors They generate sig-nals which are used for triggering the readout Thetracking station allows to change the longitudinalposition of each box and provides a specific holdingframe permitting rotations of the device under testwhich is located in the center of the telescope Theoperative setup of the tracking station is shown infigure A1

Figure A1 The Bonn tracking station set up at theCOSY accelerator (Julich)

The tracking station has been tested in differentbeam conditions

bull 295 GeVc and 893 MeVc protons at COSY(Julich)

bull 1 to 5 GeV electrons at DESY (Hamburg)

bull Bremsstrahlung photons from an electronbeam of 2 GeV at ELSA (Bonn)

bull lepton pair production from Bremsstrahlungphotons at ELSA (Bonn)

A11 Layout of the DAQ Chain

The data acquisition and reduction is done in sev-eral layers First all channels of the Si-Strip Sen-sors are time discretely sampled after amplificationand pulse shaping through APV25 front-end ASICswhen an external trigger is asserted The resultinganalog samples are multiplexed in a single mixed(analog and digital) output per front-end that issampled by an ADC running synchronous to thefront-end clock The ADCs are actually plug-incards attached to universal FPGA-Modules withVME form factor (see figure A2) [1] The FPGAconfiguration implements the following stages

bull frame decomposition

bull baseline compensation

bull pedestal tracking

bull hit-over-threshold assertion

bull cluster finding

bull FIFO buffering

bull noise statistics

bull slow-control interface controller

The resulting sparsified digitised cluster data isthen read out via the VME-Controller through stan-dard network link and stored to disk and processedfor online monitoring

A12 Structure of the Analysis Tools

The PandaRoot framework has been used to anal-yse the results of the measurements performed withthe tracking station (see [2]) A dedicated infra-structure was required to import the DAQ raw datain the framework and there perform digitisationhit reconstruction and analysis with the same toolsusually applied to simulated data The structure ofthese tools is shown in figure A3

The first step toward the final analysis is the align-ment This was done in two steps first an op-tical check by means of a laser pointer was per-formed during the installation of the modules in

146

Figure A2 Layout of the DAQ chain

Figure A3 Structure of the analysis tools

the beam lines Then an iterative off-line proce-dure based on the residuals method was applied tomeasured data in order to find the real positions ofthe sensors This information was inserted in thegeometry definition provided to the analysis frame-work A map linking front-end channel numbersto simulation-like strip identifiers was implementedThe calibration of the sensors envisages the energyloss measurements available with this kind of detec-tors The same charge was injected on each channelof the front-end electronics to resolve differences inthe response

Figure A4 Energy loss measured in one sensor withbeams of protons of two different momenta

This set of tools allowed to perform clustering andhit reconstruction inside the PandaRoot simulationframework This is a key feature to directly com-pare simulated and measured data as shown in thenext sections An example of the functionality ofthis infrastructure is shown in figure A4 where en-ergy losses obtained with protons of different mo-menta are compared

A13 Rotation of One Sensor

The tracking station is designed to permit the ro-tation of one sensor changing in this way the inci-dent angle of the beam to the detector plane (seefigure A5) This feature is important to study the

Figure A5 Top view of the setup used for the mea-surements performed rotating one double sided sensor

effect on cluster size and energy loss distributionsIncreasing the incident angle a particle will crossthe sensor along a longer path thus having in av-erage a bigger energy loss and hitting more stripsThese effects were studied during a beam test atDESY where 4 GeV electrons were casted Thereone of the double sided sensors was positioned onthe rotating holder and several incident angles were

147

set As foreseen the peak position of the energy lossdistributions moves to higher values as the incidentangle gets bigger (see figure A6) A similar trend ispredicted by simulations summarised in figure A7

Figure A6 Peak energy loss of 4 GeV electrons mea-sured in one sensor as a function of the rotation angleof the module with respect to the beam direction

Figure A7 Simulation of the measurements shown infigure A6

The cluster size is another variable influenced bythe incident angle of the beam Figure A8 showsthe average size of the clusters (expressed in termsof number of strips on both sides belonging to thesame cluster) as a function of the rotation angleAn enhancement is present for an angle of about9 Such a particular value corresponds to crossthe full width of a strip while spanning the fullthickness of the sensor1 This feature is present insimulations as well as shown in figure A9 Simu-lations present some differences in the cluster sizesalthough the trend is compatible with that one ob-tained with measured data This is due to some

Figure A8 Measured cluster size (in terms of digis)as a function of the rotation angle

Figure A9 Simulation of the effect of rotation on thecluster size

thresholds set in the MC simulations for examplethe minimum charge that can be collected by eachstrip We could not set these thresholds accordingto the values observed with the real devices sinceat the moment the absolute ADC energy loss cali-bration is not taken into account Changes of thesevalues can have a significant impact on the clustersize A too high value for this threshold can resultin a cluster with smaller multiplicity than that oneobtained from measurements On the other side toolow thresholds in MC simulations can lead to biggerclusters than the measured ones since strips wherethe collected charge is smaller than the hardwarethreshold are taken into account

1 α = arctan(

pitchthickness

)= arctan

(50microm300microm

)= 946

148

A14 Shift of One Sensor

The longitudinal position of the four boxes can bemodified in a wide range One of the sensors wasmoved along the beam axis during a beam time with3 GeV electrons at DESY (see figure A10) This

Figure A10 The figure shows how the second boxwas used to scan different longitudinal positions

scan was performed in order to understand the ef-fect of the position of the second box on the track-ing performance Therefore the unbiased resolu-tion was extracted with different setups

RESx = 4radicσx1 lowast σx2 lowast σx3 lowast σx4

where σxi is the standard deviation of the distribu-tion of the x-residuals for the ith sensor Resultsare shown in figure A11 Measurements show thatthe best unbiased resolution is obtained when thesecond box is as close as possible to the first oneThis means that the effect of multiple scattering inthe first box is not negligible at these energies andthe best condition is the one where a second pointis measured before the scattering cone diverges toomuch from the straight trajectory

A15 Resolution Estimations

The simulation framework allows to investigate theresolutions achievable with different setups A com-parison has been realised using three different se-tups and a beam of 5 GeV electrons The simula-tions were performed placing and ideal silicon sen-sor with infinite precision at the center of the track-ing station (see table A1) This ideal sensor was setas a 300 microm thick silicon volume with square areaThe position measured by this device was comparedwith the one extrapolated from the tracking stationhits This last evaluation was done calculating theintersection of the straight line fitted on the track-ing station points with the plane of the ideal sensorHistograms were filled with the x and y distance be-tween the measured and the extracted points andthe distributions were fitted with Gaussians The

Figure A11 Results of the longitudinal scan per-formed at DESY with 3 GeV electrons The secondbox was moved along the tracking station in a 50 cmwide range The figures show the estimator previouslydefined as a function of the position of the second box

B1 B2 Device B3 B4

z (cm) z (cm) z (cm) z (cm) z (cm)

A 16 86 110 145 1855

B 90 100 110 120 130

C 65 85 110 139 159

Table A1 The position of the four boxes and of theideal sensor are shown

results are shown in table A2 The best setup (B)is that one with all the sensors placed as close aspossible to the device under study This setup is notfeasible with the actual tracking station due to some

149

Setup σx σymicrom microm

A 56 53

B 16 16

C 34 34

Table A2 Results of the simulations performed withdifferent setups

limitations of the holding structures The setup Crepresents the positioning with the smallest achiev-able distances between the device and the trackingstation boxes

A16 Scattering Measurements

The tracking station has been used to measurescattering in some volumes of different materialsboth with protons at COSY and with electrons atDESY Measurements have been performed placingthe scatterers on the central holder of the track-ing station with two hits measured upstream andtwo downstream of the volume (see figure A12)The samples used for these studies were carried out

Figure A12 Setup used to measure the multiple indifferent carbon volumes One double-sided strip sensorand two single-sided ones were used both before andbehind the volumes

with materials foreseen in the implementation of theMVD support structures Both carbon and carbonfoams were characterised as well as complex proto-types for the pixel staves

bull empty volume just air between the boxes

bull 1 cm thick carbon volume with a density ofsim 179 gcm3


bull 25 cm thick of a carbon foam with a densityof sim 052 gcm3

bull a 4 mm thick pixel stave prototype realisedwith carbon foam and embedded cooling pipes(see figure 530)

The analysis focused on the determination of theprojected scattering distributions For each setupthe x and y projected scattering angle distributionswere measured (same examples are shown in fig-ure A13) In their central part these distributionare well approximated by Gaussian centered aroundzero In the following we will refer to σx and σymeaning the standard deviations obtained with aGaussian fit on the x and y projected distributionsrespectively

Figure A13 Distributions of the projected scatter-ing angles obtained with 295 GeVc protons and thepreviously described scattering volumes

The results obtained allow to distinguish the effectof different scatterers even with light ones In or-der to validate the simulation framework as a toolfor estimates of the effect of passive materials a fullset of simulations has been performed All the se-tups used in real measurements were reproduced al-lowing a direct comparison of simulations and realdata Table A3 summarises the results obtainedfrom measurements and simulations using differentscatterers and particle beams Simulations are infull agreement with the measurements performedwith protons A discrepancy of the order of some

150

Vol Ptc Mom Meas σ Sim σ

GeVc mrad mrad

1 cm C p+ 295 102 101

2 cm C p+ 295 134 133

Air eminus 10 124 140

Air eminus 30 0423 0476

Air eminus 54 0243 0284

Foam eminus 10 218 254

Stave eminus 10 176 187



Table A3 Comparison between the sigma of Gaussianfits on the distributions of the projected scattering an-gles obtained with real measurements and Geant3 sim-ulations using different scatterers

10 between measured and simulated values hasbeen found for electrons due to a stronger effectforeseen in Geant3 Anyway the scaling of scatter-ing with momenta and different scatterers is reallywell described by simulations and we can rely on thepredictions of the simulation framework to evaluatethe effect of passive materials

References

[1] R Schnell M Becker K-Th Brinkmann KKoop Th Wurschig and H-G Zaunick FPGA-based readout for double-sided silicon strip de-tectors JINST 6 (C01008) 2011

[2] S Bianco Characterization of the PANDAMicro-Vertex-Detector and Analysis of the FirstData Measured with a Tracking Station IEEENuclear Science Symposium Conference Record(N42(278))1149ndash1152 2010

151

B Julich Readout System

B1 Overview

For the development of the Micro Vertex Detector(MVD) the evaluation of prototypes and detectorparts is very important Different prototypes of thepixel front-end chip ToPix (Torino Pixel) need tobe tested and characterised under similar conditionsto improve the development

To control these devices under test (DUT) and tosave the taken data a suitable readout system isnecessary To have similar conditions for differentprototypes and development stages a modular con-cept of a readout system is required which can beadapted in a simple way to the specific interfaceof different types of electronics At the same timethe system must provide a high performance to al-low the evaluation of single front-end chips as wellas fully assembled modules The possibility to im-plement routines for online data processing is alsodesirable

A digital readout system has been developed tomeet all these requirements for the development ofthe MVD the Julich Readout System

Figure B1 Schematic drawing of the Julich ReadoutSystem

The Julich Readout System is a compact and pow-erful table top setup which allows a quick testingof new detector components The setup consists of(see figure B1)

bull Digital readout board

bull FPGA Firmware

bull MRF Software with GUI

bull Readout PC

bull Device under test

bull Adapter board PCB

The central component is the FPGA1 based read-out board It is connected to the readout PC viaan optical connection which is used to receive com-mands from the user and to send data to the PC forfurther processing The corresponding software in-frastructure is implemented within the MVD read-out framework (MRF) The prototype is controlledby the user with a graphical user interface (GUI)On the other side the readout board is connectedto the DUT on its test board An intermediateadapter board converts the DUT signal interface tothe interface of the readout board

B2 Basic Concepts

In the following we will outline the basic concept ofthe Julich Readout System and introduce the indi-vidual components comprising the modular design

Modularity

Complex procedures are broken down into simplersubtasks which are each handled by individual mod-ules To add or change functionality it is only nec-essary to add a new module or to change an existingone limiting required changes to individual moduleswhich significantly improves stability and simplifiesdebugging The modules can also easily be reusedwhen already implemented subtasks reappear in adifferent context This modular concept makes thereadout system very flexible and easy to maintain

Communication Layers

In addition to the modular setup the MRF followsthe principle of different abstraction layers basedon the OSI model2 [2] Following this principlethe MRF defines four communication layers (see fig-ure B2) Every hardware component of the read-out system corresponds to a specific communica-tion layer and has a representation in software orfirmware Each layer uses the functionality of the

1 Field Programmable Gate Array

2 Open Systems Interconnection model

152

Figure B2 Schematic overview of the Julich Read-out System Every hardware component of the systemcorresponds to one of the four defined communicationlayers and has a representation in software or firmware(after [1])

layer directly below without knowledge of the exactimplementation and provides a well-defined set offunctions to the layer above Due to this approachlayers can be replaced without changes to the otherlayers The MRF defines the following layers

bull Layer 1 - Physical Layer Establishing of signalconnections

bull Layer 2 - GAL (Generic Access Layer) Com-munication with readout board via SIS1100 op-tical connection

bull Layer 3 - TAL (Transport Access Layer) Ac-cess to functionality of the readout board

bull Layer 4 - CAL (Chip Access Layer) Commu-nication with DUT

In the following we list the most important compo-nents of the readout system and indicate to whichcommunication layer they belong

Adapter Board Layer 1

The adapter board establishes the signal connec-tion between the readout board and the DUT Dueto the usually non-standard interface of the DUTtestboard a specific adapter board needs to be de-veloped for each type of DUT

Digital Readout Board with Firmware Layer 1and 3

The FPGA based digital readout board is the mainhardware component of the system The VHDL3

based firmware configures the FPGA with the im-plemented functionality (layer 3) and the pin as-signment (layer 1) of the DUT interface

Optical Connection to Readout Board Layer 2

The data transport from the digital readout boardto the PC is done via a 1 Gbits optical connection(SIS1100 [3]) On the PC side a SIS1100 PCI cardis installed The SIS1100 interface of the digitalreadout board is implemented as a SIS1100 CMC(Board A) or as a combination of a SIS1100 FPGAcore and an SFP4 optical gigabit transceiver(Board B)

MRF Readout Software Layer 2 3 4

The MRF readout software (short MRF) is themain software component of the readout systemHere the communication layers 2 3 and 4 are de-fined (see chapter B24) These layers establishconnections to the DUT and the firmware of thedigital readout board

MRF-Control Graphic User Interface

A graphical user interface (GUI) has been devel-oped in C++ based on the Qt framework (see [4])for an easy control by the user Full configurationand readout routines are available Measured datacan be easily saved to disk The settings of theDUT and the setup can be saved and reload in spe-cial configuration files

B21 Digital Readout Board A

The first digital readout board (see figure B3) isa custom development and production of the Zen-trallabor fur Elektronik (ZEL Central Laboratoryfor Electronics) of the Forschungszentrum Julich[5] A Xilinx Virtex 4 FPGA configured with thefirmware takes the control of the DUT communica-tion and of the data buffering The readout boardis connected to the PC via a 1 GBits optical link

3 Very High Speed Integrated Circuit HardwareDescription Language

4 Small Form-factor Pluggable

153

Figure B3 The digital readout board A of the Read-out System with SIS1100 CMC (left) Xilinx Virtex 4FPGA (middle) power connectors (right) and 68-pinconnector to adapter board (middle right) (from [1])

An optical fiber has the advantage of galvanic sepa-ration of PC and readout board The SIS1100 pro-tocol is used to transfer the data from the bufferto the PC To implement the SIS1100 protocol onthe readout board A a Common Mezzanine Card(CMC) with SIS1100 functionality is connected tothe readout board The adapter board is connectedby a 68-pin flat band cable

The main features and external interfaces of thedigital readout board are

bull Xilinx Virtex 4 XC4VLX60 FPGA

bull 32 MB Xilinx Platform Flash for firmware stor-age

bull 64 free configurable FPGA inputoutput pads

bull 16 LVDS inputs

bull 16 LVDS outputs

bull SIS1100 based optical connection to PC with1 Gbits

bull 50 MHz single ended clock

B22 Digital Readout Board B

To meet the requirements of an upcoming full sizeToPix prototype and online analysis an upgradeof the digital readout board was developed TheML605 evaluation board is a commercial productby Xilinx and is comparatively cheap and easy to

get It contains a DDR3 RAM slot for data storageand a Virtex 6 FPGA (see figure B4) The con-nection to the PC is established via a SFP gigabittransceiver The SIS1100 protocol is implementedas a IP core into the firmware

Specification of the ML605 are

bull Xilinx Virtex 6 XC6VLX240T FPGA

bull DDR3 SODIMM memory (expandable)

bull SFP Module connector

bull FMC connectors for Standard MezzanineCards

bull LC-Display

bull Compact Flash Card for firmware storage

bull 66 MHz single ended and 200 MHz differentialclock

B23 Firmware

The FPGA on the digital readout board is config-ured with a firmware which implements the desiredfunctionality and configures the external interfaces

The firmware for both readout boards is written inVHDL a hardware description language and thensynthesised with the Xilinx development tools ietranslated to a format which is understandable bythe FPGA

The firmware is divided into modules which handledifferent subtask concerning the DUT the boardfunctionality or the communication All modulesare connected to the register manager which is itselfconnected to the SIS1100 interface (see figure B5)

Figure B4 The digital readout board B ML605 fromXilinx with a Virtex 6 (middle) SFP port (up left)FMC connectors to adapter board (up middle) Com-pact flash slot (right) and DDR3 RAM (next to FPGAon the left)

154

Data which arrive via the SIS1100 are formattedin address-data pairs The register manager willdistribute the data depending on the address to theconcerning modules If data should be read fromthe board the register manager will take the datafrom the module indicated by its address

This modular design makes it easy to adapt thefirmware to a different readout board or a differentDUT respectively

Figure B5 Schematic firmware setup Data deliveredvia SIS1100 will be distributed to the modules indicatedby the address The red arrows indicate the path of theshown example data packet

B24 MVD Readout Framework

The MRF software framework implements the up-per three levels of the communication model TheMRF is written in C++ and has no external de-pendencies except the Standard Template Library(STL) to avoid conflicts with other libraries Thismakes the MRF software a decoupled packagewhich can be used in any environment

For every communication layer a basic set of com-mands is implemented to provide the data transportbetween the layers In addition specific commandsfor the used hardware are implemented

GAL Generic Access Layer

The generic access layer provides the basic func-tionality for communication with the digital read-out board The read and write commands provide abasic reading and writing to registers over an openconnection

All more complex commands of the upper layers areimplemented using these basic functions

Including the SIS1100 driver into the GAL morepowerful functionality is implemented TheSIS1100 driver provides DMA transfer5 Data froma buffer on the readout board can be send directlyto a data container structure in the MRF software

In addition the readout board is able to send in-terrupt signals to the PC to indicate certain condi-tions such as data from the DUT being available Atransfer to the PC can then be triggered automati-cally

TAL Transport Access Layer

The transport access layer provides access to thefunctionality of the firmware For both versions ofthe digital readout board certain basic functionalityis available such as the online configuration of theclock generator and the controlling of status LEDs

CAL Chip Access Layer

The Chip Access Layer implements the functional-ity of the DUT and needs to be adapted to the spe-cific hardware eg configuration of the chip or thereadout sequence Two versions have been made forthe control of the Atlas FE-I3 and another one forthe ToPix v2 The complete readout and all testingprocedures are available for both chips

B3 Conclusion

The Julich Readout System has been used to testthe front-end prototypes ToPix v2 for the MVDand FE-I3 as a fallback solution The FE-I3 hasbeen tested with the digital readout board A theToPix v2 with both the readout boards A and BDue to the modular design of the digital readoutsystem the adaption to the different chips and hard-ware was very fast Different measurements like sta-bility tests with different clock frequencies thresh-old scans and tuning of threshold dispersions havesuccessfully been done [6] [1]

References

[1] Marius C Mertens Der PANDA Mikro Ver-tex Detektor Endwicklung eines Labormessys-tems Simulation der MVD-Betriebsparameter

5 Direct Memory Access

155

sowie Untersuchung zur Auflosung der Breitedes Dlowasts0(2317) PhD thesis Ruhr-UniversitatBochum 2010

[2] H Zimmermann OSI reference model - the ISOModel of Architecture for Open Systems Inter-connection IEEE Transactions on Communi-cations 28(4)425ndash432 1980

[3] M Drochner et al A VME controller fordata aquisition with flexible gigabit data link toPCI Proceedings of the 12th IEEE Real Timecongress on Nuclear and Plasma Science June2001

[4] Qt 4 Framework wwwtrolltechcom last vis-ited on Nov 22 2011

[5] H Kleines et al Developments for the Readoutof the PANDA Micro-Vertex-Detector IEEEConference Record NSS-MIC-RTSD 2009

[6] David-Leon Pohl Charakterisierung einesSilizium-Pixel-Auslesechips fur den PANDAMikro-Vertex-Detector Masterrsquos thesis Univer-sitat Bochum 2011

156

157

C Details on Vertexing

In section 672 a characterisation of the vertexingperformance was shown Four pions were propa-gated though the MVD volume from different com-mon start vertices and the achieved resolutions weremapped as a function of the position of the vertexAll the scans showed a different resolution for thex and y coordinates of the reconstructed verticesWhen the polar angle (θ) of the pions reaches valuesat around 50 the vertex resolution for the x coor-dinate becomes worse than the y one This is notexpected because of the φ symmetry of the MVDdesign In order to find out the reason of this ef-fect several tests were performed First of all thevertex finders have been characterised in detail tofind out possible critical behavior Table C1 showsthe effect of an artificial rotation of 90 applied totracks in the x-y plane before applying the ver-texing This test was performed with four pions of1 GeVc momentum distributed over the full angularranges (φ isin [0 360] and θ isin [10 140]) The ro-tation moves the worse reconstruction performanceto the y coordinate excluding a bias introduced bya different treatment of x and y by the vertexingtools Several investigations lead to the conclusion

σ (microm) No Rot 90

x 883 694

y 694 883

Table C1 Results obtained with an artificial rotationof track candidates before the vertex finding (POCA)

that this effect is due to the constrains to the designof the inner barrel layers Because of the presenceof the target pipe and due to mechanical restric-tions it was impossible to cover with pixel sensorsthe full φ acceptance in the area immediately down-stream of the interaction point (a schematic view isshown in figure C1) The design of barrel 1 and2 does not foresee ldquoforwardrdquo barrel sensors in thetop and bottom regions This translates into thefact that tracks flying into these holes in the accep-tance create a smaller number of hits in the MVDand have bigger distances between their first hitand the interaction point (see figure 619 and fig-ure 621) These particles will be reconstructed withworse performances leading to worse vertex deter-mination Since these critical regions are in the topand bottom regions of the MVD they effect muchmore the determination of the x coordinate of thevertex than y This appears clearly if one propa-

Figure C1 Schematic side view of the barrel sensorscovering the φ regions around the target pipe

Figure C2 Configurations used for the simulations

gates four pions from the interaction point settingtheir φ angles so that they are in an orthogonaldisposition in the transverse plane There one canrotate differently the set of tracks and check the ef-fect on vertexing performances Figure C2 showsthe two configurations chosen to investigate the ef-fect of the gaps in case ldquoArdquo the four pions do notenter the target pipe and they all create a hit inthe first barrel layer while in case ldquoBrdquo two tracksare propagated with initial directions pointing in-side the target pipe hole In both cases θ was setto 75 for all the tracks Table C2 summarises theresults obtained with the configurations A and Bboth with and without the material of beam andtarget pipes defined in the geometry It clearly ap-pears that setup B produces a much worse vertexingperformance for the x coordinate while the configu-

158

Setup σx(microm) σy(microm)

A cross no pipes 30 30

A cross pipes in 33 33

B cross no pipes 103 29

B cross pipes in 103 30

All φ pipes in 67 47

Table C2 Resolutions obtained with the different se-tups and kinematics θ was fixed to 75

ration A shows the same behaviour for x and y Forthese last studies a different vertexing tool has beenused (the Kinematic Vertex Fit) to cross-check thatthe x-y different performances were not induced bythe vertexing itself This is the reason why resultshere are different from the ones obtained in the restof this section (there the POCA method was used)The different x-y resolutions appeared to be strictlyrelated to the polar angle of the tracks For parti-cles flying in the forward region the performanceis the same for x and y At θ of about 50 thedifference starts to appear This can be seen infigure 628 where the x and y resolutions obtainedwith different θ values (integrating over the full φ)are summarised

159

D Alignment

D1 Introduction

In track reconstruction and physics analysis the de-tector alignment is a mandatory task An accuratedetermination of a large number of parameters isrequired to allow precise track and vertex recon-struction

The momentum resolution of reconstructed tracksdepends in part on the alignment of the detectors inspace It should not be degraded significantly (com-monly 20) with respect to the resolution expectedin case of the ideal geometry without misalignment

This implies that the detectors positions must beknown to an accuracy better than 20 of the in-trinsic resolution This corresponds to an alignmentprecision of 10 microm or better in the bending plane

Optical surveys and full mapping of MVD modulesand super-modules can be a very useful startingpoint and a constraint in the alignment procedurebut do not provide the required accuracy and can-not correct the time dependent changes of the geom-etry after the tracker installation Module positionfluctuations (O(10 microm)) and barrelstavedisk de-formation (O(100 microm)) can occurr Therefore align-ment algorithms based on the track reconstructionshould be used for the determination of the sensorpositions and orientations The alignment proce-dure aims to determine the positions of the modulesof each sub-detector The basic problem is sketchedin figure D1

Figure D1 The basic alignment problem the correctmodule geometry (left) must be deduced by studyingthe deviation of the hit positions (red points) with re-spect to the tracks

The barrels and the disks require different strategiesand the last is the more difficult task The best can-didates for barrel layers alignment are cosmic raysof high momentum(p gt 3 GeVc) with or withoutmagnetic field The availability of collision tracksamples will be investigated in particular to alignthe MVD disks since they are less often traversedby cosmic ray tracks

D2 General Strategies andTechniques forVertexTracking Detectors

Any tracking system can be considered as an assem-bly of separate modules whose position can be dis-placed during assembling during detector mainte-nance and during data taking periods as well due topossible varying environmental conditions (temper-ature humidity elastic strains gravitational sag-gings etc) Therefore alignment procedures areneeded to detect and correct possible distortionsAny deformation of the mechanical structure of themodules or their mislocation from the nominal po-sition implies a depletion of the particle positiondetermination Therefore mapping and recoveringthe position of the any tracking detector (especiallythe vertex ones) is of central importance to preservethe intrinsic spatial resolution of the detector dur-ing a long data taking

The alignment of large detectors in particle physicsusually requires the determination and tuning of alarge number of alignment parameters typically notsmaller than 1000 Alignment parameters for in-stance define the space coordinates and orientationof detector components Since they are commonto all data sets of real measurements produced bythe tracking system they are called global parame-ters Usually alignment parameters are correctionsto default values (coming from a topological surveyfollowing the detector installation) so they are ex-pected to be small and the value zero is their initialapproximation Alignment parameters are usuallylocally defined ie in each module reference system

In a three-dimensional space the most general setof alignment parameters able to describe any dis-placement is constituted by 12 figures

bull three rescaling of coordinate axes

160

bull three global traslations

bull three rotations

bull three shearings

The four basic types of linear displacements areshown in figure D2

Under certain conditions some of these parametersmay be fixed and the number of global parame-ters can be reduced For instance for a stack ofrigid modules disposed vertically with respect to thebeam axis (z as can be a stack of forward moduli)the rescaling along x and y as well as a xy shear-ing would imply a deformation of the modules andof their pitch and so these degrees of freedom canbe fixed In this particular geometrical layout alsoshearings and rotations basically coincide

Alignment procedures are usually data-driven sothat displacements are determined comparing themeasured values of hits by tracks from physics in-teractions and from cosmics and fitted track co-ordinates The deviations between the fitted andthe measured data the residuals are used to deter-mine the alignment parameters by means of properleast squares fits Tracks are also defined througha series of parameters which for instance describetheir slopes or curvatures and are valid for each sin-gle track only They are called local parameters todistinguish them from the global alignment onesDespite their locality track parameters must be de-fined in the global (ie the whole detector) refer-ence system Together with alignment parametersthey define the complete solution of the minimisa-tion procedure

In general the alignment strategy of a stack oftracking modules proceeds through a series ofstages which can be shortly summarised as follows

1 Precision Mechanical Assembly Themodules are accurately assembled andmounted onto high precision base supportsThe mounting tolerances define the scope ofthe possible misalignment for instance if amodule can be placed in a box with a 20 micromaccuracy one can expect misalignements tobe distributed as a Gaussian centered on zeroand 20 microm wide

2 System Metrology At the time of construc-tion coordinated measuring machines must beused to survey and fully map the surfaces ofindividual modules Also mounted frames co-ordinates have to be fully surveyed in stan-dard conditions of temperature pressure andhumidity Results from survey have then to

be compared to nominal designs fixing macro-scopic deviations The outcomes from the me-chanical surveys will provide the reference po-sitions of all components

3 Software Alignment If the quality of me-chanical survey is good enough alignment maybe performed offline during data taking oreven afterwards Otherwise a fast alignmentengine must be available to tune-up the start-ing alignment configuration before data takingand possibly run whenever critical conditionsare met In general the software alignmentmust be flexible enough to account for a widerange of possible occurrences arising during thedata taking Statistical methods are used to re-cover the actual position of the modules as willbe described in the following

4 Alignment Monitoring A monitoring ofthe quality of detector alignments during datataking can sometimes be advisable a monitor-ing deamon checking periodically the qualityof track residuals during data taking could au-tomatically trigger the re-running of softwarealignment procedures to tune online the align-ment quality

In the following the Software Alignment stages willbe described in some detail presenting the availabletools commonly used in the alignment proceduresof some particle physics vertex detectors

D21 Software Alignment of TrackingModules and Modern Tools

The analysis of the residual distributions and theirminimisation allow to position detector modules intheir actual locations with an accuracy which is onlylimited by the available statistics of tracks to befitted and the detector intrinsic resolution

Track residuals are functionally dependent on theboth global and local parameters The problem ofalignment is in general formulated as a mathemat-ical problem of minimisation of the functional ob-tained by the squares of residuals of both global andlocal parameters

χ2 =sum

events

sumtracks

sumhits

∆2i σ

2i (D1)

where ∆i = xfitminusxmeas is the hit residual and σi theaccuracy of the hit coordinate measurement Thefitted track position xfit has as well a functional de-pendence both of the geometrical misalignment ofthe module and of the track parameters A track

161

Figure D2 Schematic view of the four basic types of linear transformations

model must be chosen to define the track depen-dence on the free parameters For instance in thesimplest linear model the observables have expec-tation values which are linearly dependent on theset of parameters In this case the fitted trackposition may be given by the linear sum of twocontributions one pertaining to the n track localparameters ai and one to the m global alignmentones αj together with their derivatives di and δjrespectively

xfit =

nsumi=1

aidi +

νsumj=1

αjδj (D2)

Using a linear track model the set A = ai+αiof parameters minimising the χ2 quadratic func-

tional is obtained solving the set of normal linearequations of least squares in matrix form C = AbC is a symmetric N timesN matrix (N = n+m) andb a N -vector both of them being given by sums ofindependent contributions from each measurementThe solution of this system of linear equations posesthe major problem of the alignment task since thenumber of parameters to be determined (ie the di-mension N of the C matrix of the normal equationsto be inverted to find the A covariance matrix) isat least of the order of 103 This large parame-ter multiplicity requires a number of tracks to befitted by the alignment procedure of the order of104 minus 106 to get a reasonable accuracy Howeverthis is not a severe complication as C and b already

162

contain the complete information from single mea-surements which thus are not needed to be storedany further A practical limit to the number ofparameters is given by the time necessary to per-form the matrix inversion which is proportional toN3 and to the storagememory space (prop N2 sincedouble precision calculations are often mandatory)This means that a number of parameters larger than104 starts being critical if the matrix equation is tobe solved by standard methods With 10000 param-eters a typical computing time for matrix inversionis about six hours which becomes almost one yearfor N = 105

For the PANDA MVD the typical number of ex-pected degrees of freedom (for pixels stations andstrips) is about 5000 about half as much as LHCexperiments Thus at least execution times shouldnot represent a problem

However the normal equation matrix could easilybecome singular in case of missing data (for in-stance dead channels) or in case of a completecorrelation between parameters (for instance whentracks are projected into themselves by geometricaltransformations) In this case the matrix inversionfails and the analytical solution of the problem isprevented This requires alternative approximateor iterative approaches or special numerical algo-rithms to cope with the problem

A first rough approximation is to perform a seriesof track fits which initially neglect alignment pa-rameters in this case the fits depend only on thereduced set of local parameters and are easier toperform However this method is not statisticallycorrect since the track model is basically wrong andthe obtained results are biased and possibly alsounstable as affected by detector inefficiencies nottaken into account Through several iterations themethod can effectively reduce the residuals (thoughnot necessarily converging) but the alignment pa-rameters applied as corrections in iterative fits willstill be biased Moreover the correlations betweenlocal and global fits present in the functional inthis way are completely ignored

A better solution to reduce the problem dimensionis to use tracks extrapolated from outer detectorsin order to fix at least the parameters of thet tracksIn this case one can align a given subdetector (forinstance stack of modules along a given axis or ina detector sector) with respect to other parts of thesystem The method has the advantage of provid-ing the alignment of the detector as a whole withhigh precision On the other hand it works effec-tively only when the detector dimensions are nottoo large because with large extrapolation lenghts

the precision of the extrapolated track parameterscan soon become much worse than that for tracksreconstructed in the subdetectors only due to mul-tiple scattering effects and misalignements of thereference detector out of control This is actuallythe method adopted for the first stage of alignmentof the FINUDA vertex detector [1] composed bytwo layers of double-sided silicon microstrip mod-ules (50 microm pitch) arranged in cylidrical geometryaround the beam axis The applied procedure wasbased on a minimisation engine which aligned bymeans of straight cosmic rays stacks of four mod-ules at a time two per each of the two detectorlayers (starting from those crossed by a larger num-ber of straight cosmic rays) Then adjacent sin-gle modules of the innerouter layers were alignedone at a time with respect to the already alignedstacks The error in the alignment of a single mod-ule assuming multiple scattering effects are negligi-ble is given by the sum in quadrature of the mod-ulus intrinsic resolution (σintr) and the alignmentprecision (σalign) The minimum number of tracksneeded to perform a reliable minimisation is deter-mined by the condition (σalignσintr)

2 1 Once asatisfactory relative alignment of the two layers wasachieved they were then aligned with respect to theouter tracker (a straw tube array) which could pro-vide information on global rotations and tilts of thevertex detector as a whole With this method afterhaving taken into account multiple scattering andsagging effects a resolution of the vertex detectormoduli of about 20 microm could be achieved for the φcoordinate and better than 30 microm for z For thesake of comparison the starting mechanical toler-ance of the holder devices was sim 200 microm

The most efficient alignment method consists ofcourse in making a simultaneous fit of all global andlocal parameters by means of dedicated algorithmswhich could overtake the above mentioned matrixinversion problem To this purpose the MILLE-PEDE II programme was developed [2 3 4] and it ispresently widely used for the alignment of large par-ticle physics detectors (among them ALICE ITS[5] LHCb VELO [6] HERA-B OTR [7 8] and H1SVD [9]) Also FINUDA used it for fine tuning pur-poses) The MILLEPEDE II programme is based onpartition techniques which split the problem of in-version of a large matrix into a set of smaller in-versions exploiting the feature of the C matrix ofbeing very sparse presenting many null submatriceswhose computation can be skipped reducing execu-tion times Depending on the fraction of vanish-ing submatrix elements alignment problems with anumber of global parameters of the order of 105 canbe attacked with standard computing power Of

163

course the matrix size ought to be conveniently re-duced whenever possible before starting the min-imisation procedure To this purpose some prelim-inary iterative steps are needed in order to

bull linearise the equations (by means of Taylor ex-pansions) if the dependence on the global pa-rameters happens not to be linear and

bull remove outliers (ie data with large residuals)from the data by proper cuts (to get tighter inseveral iterations) of properly down-weightingthem

Sometimes the minimisation is not sufficient toachieve an optimal alignment if some degrees offreedom are undefined which call from additionalconditions expressed by linear (or linearised) con-straints The MILLEPEDE II programme can han-dle additional these linear constraints (and param-eters) exploiting the Lagrange multiplier method

The MILLEPEDE II package is originally built bytwo parts The first (MILLE) is called in user pro-grams and basically prepares the data files (con-taining measurements and derivatives with respectto local and global parameters in their proper ref-erence systems) These files are processed by thesecond standalone program (PEDE) which per-forms the fits and determines the global param-eters steering the solutions The solutions forlocal parameters which are track dependent arenot needed The package originally written inFORTRAN-77 is available for FORTRAN and C-based aligment procedures and can be easily em-bedded in existing alignment frameworks (AliMille-Pede in AliRoot LHCb VELO alignment softwarepackage [10 11 12])

The MILLEPEDE II package allows to achieve verysatisfactory alignment conditions Using this frame-work the forward stations of the LHCb VELO de-tector for instance could be aligned with a preci-sion of 3 microm on x and y translational coordinates (tobe compared with an intrisic resolution larger than5 microm) and with an accuracy of 04 mrad on therotational degrees of freedom (around the z axis)The alignment of this detector was performed usingtracks from LHC pp collisions and tracks from beamhalo particles crossing the whole detector Suchan approach can also be envisaged for the PANDAMVD forward disks

References

[1] S Piano Search for Sigma Hypernuclear stateswith the FINUDA Spectrometer PhD thesisUniversita degli Studi di Trieste 2001

[2] V Blobel Software alignment for tracking de-tectors Nucl Instr Meth A566 2006

[3] V Blobel and C Kleinwort A new methodfor the high-precision alignment of track detec-tors Number DESY-02-077 hep-ex0208021March 2002

[4] V Blobel MILLEPEDE II package descriptionand code httpwwwdesyde~blobel

[5] ALICE Collaboration JINST 5 (P03003)2010

[6] LHCb Collaboration LHCb VELO technicaldesign report LHCb TDR 5 (CERN-LHCC2001-010) 2001

[7] D Krucker OTR alignment method descrip-tion www-hera-bdesyde~krueckerOTR

html

[8] R Mankel A Canonical Procedure to FixExternal Degrees of Freedom in the InternalAlignment of a Tracking System HERA-B 99-087

[9] D Pitzl et al The H1 silicon vertex detectorNucl Instr Meth A454 2000

[10] S Viret C Parkes and D Petrie LHCb VELOsoftware alignment ndash Part I the alignment ofthe VELO modules in their boxes (GLAS-PPE2006-03) 2006

[11] S Viret C Parkes and M Gersabeck LHCbVELO software alignment ndash Part II the align-ment of the VELO detector halves (GLAS-PPE2007-13) 2007

[12] S Viret M Gersabeck and C Parkes VELOalignment webpage ppewwwphglaacuk

LHCbVeloAlign last visited on Nov 22 2011

164

165

3D 3-Dimensional

ADC Analog to Digital Converter

ALICE A Large Ion Collider Experiment

ALICE ITS ALICE Inner Tracking System

APV Analog Pipeline Voltage mode

ASIC Application Specific IntegratedCircuit

ATCA Advanced TelecommunicationsComputing Architecture

ATLAS A Toroidal LHC ApparatuS

BERT Bit Error Rate Test

BTeV B meson TeV


CCD Charge-Coupled Device

CCU Chip Control Unit

CERN Conseil Europeen pour laRecherche Nucleaire

CMC Common Mezzanine Card

CMOS Complementary Metal OxideSemiconductor

CMP Chemical Mechanical Polishing

CMS Compact Muon Solenoid

CNT Counter

COSY COoler SYnchrotron

CR Configuration Register

CRC Cyclic Redundant Check

CRCU Column Readout Control Unit

CSA Charge Sensitive Amplifier

CT Central Tracker

Cz Czochralski

DAC Digital-Analog Converter

DAFNE Double Annular ring For NiceExperiments

DAQ Data Acquisition

DC Direct Current

DC-DC DC-to-DC Converter

DCS Detector Control System

DESY Deutsches ElektronenSYnchrotron

DIRC Detector for Internally ReflectedCherenkov Light

DMA Direct Memory Access

DPM Dual Parton Model

DVCS Deeply Virtual ComptonScattering

ECC Error Correction Code

EDM Electrical Discharge Machining

EMC Electromagnetic Calorimeter

ENC Equivalent Noise Charge

EPICS Experimental Physics andIndustrial Control System

FAIR Facility for Antiproton and IonResearch

FBK - ITC Fondazione Bruno Kessler -Istituto Trentino di Cultura

FE Front-End

FEE Front-End Electronics

FEM Finite Element Method

FET Field-Effect Transistor

FIFO First In First Out

FINUDA FIsica NUcleare a DAFNE

FNAL Fermi National AcceleratorLaboratory

FPGA Field Programmable Gate Array

FS Forward Spectrometer

FTS Forward Tracking System

FZ Floating Zone

FZJ Forschungszentrum Julich


GBLD GigaBit Laser Driver

GBT GigaBit

GBTX GigaBit Transceiver

GBTIA GigaBit TransImpedanceAmplifier

166

GBT-SCA GigaBit Slow Control ASIC

GEM Gas Electron Multiplier

GR Guard Ring

GSI Gesellschaft furSchwerionenforschnung

GUI Graphical User Interface

H1 - SVD H1 Silicon Vertex Detector

HERA Hadron-Elektron-Ring-Anlage

HERA-B OTR HERA-B Outer TRacker

HESR High Energy Storage Ring

HL High Luminosity (HESR operationmode)

HR High Resolution (HESR operationmode)

HR High Resistivity (wafers)

HV High Voltage

I2C Inter-Integrated Circuit

INFN Istituto Nazionale di FisicaNucleare

IP Interaction Point

ITME Institute of Electronic MaterialsTechnology

LD Laser Driver

LED Light Emission Diode

LENA Laboratorio Energia NucleareApplicata

LHC Large Hadron Collider

LHCb Large Hadron Collider beauty

LHCb VELO LHCb VErtex LOcator

LNL Laboratori Nazionali di Legnaro

LR Low Resistivity (wafers)

LVDS Low-Voltage Differential Signaling

MC Monte-Carlo

MCP PMT Multi-Channel Plate PMT

MDC Module Data Concentrator

MDT Mini Drift Tubes

MicroTCA Micro TelecommunicationsComputing Architecture

MIP Minimum Ionizing Particle

MMB MVD Multiplexer Board

MOS Metal Oxide Semiconductor

MPV Most Probable Value

MR Middle Resistivity (wafers)

MRF MVD Readout Framework

MVD Micro Vertex Detector

NIEL Non-Ionizing Energy Loss

OSI Open Systems Interconnection

PANDA antiProton ANnihilation atDArmstadt

PCB Printed Circuit Board

PCI Peripheral ComponentInterconnect

PCIe Peripheral ComponentInterconnect Express

pdf Probability Density Function

PD Photo Diode

PDG Particle Data Group

PID Particle Identification

PLL Phase Locked Loop

PMT Photomultiplier

POCA Point of Closest Approach

PRBS Pseudo Random Bit Sequence

PVD Physical Vapour Deposition

QCD Quantum Chromo Dynamics

RF Radio Frequency

RICH Ring Imaging Cherenkov Counter

RMS Root Mean Square

SDS Silicon Detector Software

SEU Single Event Upset

SFP Small Form-factor Pluggable

SIMS Secondary Ion Mass Spectrometry

167

SLHC Super Large Hadron Collider

SLVS Scalable Low Voltage Signaling

SMD Surface Mount Device

SODA Synchronisation Of DataAcquisition

SP Separation Power

SR Shift Register

STI Shallow Trench Insulation

STT Straw Tube Tracker


TIA TransImpedance Amplifier

TMR Triple Modular Redundancy

TOF Time-of-Flight

ToPix Torino Pixel

ToT Time-over-Threshold

TRx Transceiver

TS Target Spectrometer

UI Unit Interval

UMC United MicroelectronicsCorporation

UrQMD Ultra-relativistic QuantumMolecular Dynamic

VCSEL Vertical-Cavity Surface-EmittingLaser

VHSIC Very-High-Speed IntegratedCircuits

VHDL VHSIC Hardware DescriptionLanguage

VL Versatile Link

VMC Virtual Monte-Carlo

VME Versa Module Eurocard

WASA Wide Angle Shower Apparatus

ZEL Zentrallabor fur Elektronik at FZJ

168

169

List of Figures

11 Overview of the future FAIR facility 1

12 Layout of the PANDA detector 2

13 Layout of the High Energy StorageRing HESR 4

14 Optical functions of HESR lattice forγtr = 62 4

15 Summary of the different target op-tions foreseen at PANDA 6

16 Time dependent macroscopic lumi-nosity profile L(t) in one operationcycle 8

17 Maximum average luminosityvs atomic charge Z of the target forthree different beam momenta 9

18 Basic detection concept 11

19 Artistic side view of the Target Spec-trometer (TS) of PANDA 12

110 The Micro Vertex Detector (MVD)of the Target Spectrometer 12

111 Straw Tube Tracker (STT) of theTarget Spectrometer 13

112 Artistic side view of the ForwardSpectrometer (FS) of PANDA 15

113 Double layer of straw tubes in thetracker of the Forward Spectrometer 16

114 Overview of the PANDA trackingsystem 18

21 Expected particle distribution inantiproton-proton collisions at differ-ent beam momenta 25

22 Expected particle distribution inantiproton-nucleus collisions at max-imum beam momenta 25

23 Positioning of the MVD inside theTarget Spectrometer and zoom of thebeam-target geometry 27

24 Basic layout of the MVD 28

25 Hierarchical structure of the MVD 29

26 Schematics of the basic pixel sensorgeometry 29

27 Basic sensor geometry for the MVDstrip part 30

28 Scattering effects in silicon and opti-misation of the stereo angle for thetrapezoidal sensors 30

29 Overall routing concept for the MVD 32

210 Overall detector integration of theMVD 32

31 Scheme of the hybrid pixel for PANDA 35

32 Leakage current as a function of thebias voltage for assemblies Epi-50and Epi-75 36

33 Threshold values corresponding tothe Landau most probable value forthe different epitaxial layer thicknesses 37

34 Annealing behaviour of the full de-pletion voltages as a function of theannealing time for the irradiated diodes 38

35 Picture of a PANDA wafer 39

36 Picture showing a partial view of thesensor wafer 39

37 Photograph of a part of the pixel ma-trix 39

38 Plot showing the unplanarity of athinned wafer prototype measuredwith a stylus probe 40

39 Oxygen concentration as a functionof wafer depth for different epitaxialthicknesses 40

310 Oxygen concentration as a functionof FZ wafer depth for some oxidationprocesses 41

311 Oxygen concentration as a functionof epitaxial wafer depth 41

312 Scheme of the wafer with the first de-sign of the full size sensors 41

313 Pixel readout scheme 43

314 Analog readout channel 44

315 Pixel control logic 44

316 Column readout schematic 45

317 Frame structure 46

318 ToPix v1 simplified schematic 47

319 ToPix v1 cell schematic 47

170

320 Cell output for an input charge of05 fC 48


322 ToPix v2 die 49

323 Dice cell schematic 49

324 Diagram of the data acquisition sys-tem for testing ToPix v2 49

325 Input-output transfer function 50

326 Deviation from linear fit 50

327 Average baseline during irradiation 50

328 Average noise during irradiation 50

329 Average baseline during annealing 50

330 Average noise during annealing 50

331 Average slope during irradiation 51

332 Tail slope sigma during irradiation 51

333 Weibull fit to the experimental iondata of the ToPix v2 configurationregister 52

334 ToPix v3 layout 45 mmtimes 4 mm die 53

335 Analog Cell Layout 54

336 Charge Sensitive Amplifier schematic 54

337 Triple well graphical representation 54

338 CSA with a feedback circuitry whichautomatically compensates for thedetector leakage current 55

339 Schematic of the constant feedbackcurrent generator 55

340 Comparator schematic 56

341 Simplified model of the 5 bit DAC 56

342 Calibration circuit schematic 56

343 Reconstructed signal shapes withthreshold scan for different inputcharge values 57

344 ToPix v3 extrapolated baseline be-fore and after equalisation 57

345 ToPix v3 baseline distribution beforeand after correction 57

346 ToPix v3 pixel map for the fourtested samples 58

347 ToPix v3 measured analogue gain 58

348 ToPix v3 measured noise 58

349 ToPix v3 transfer function for nomi-nal feedback current - full range 58

350 ToPix v3 transfer function for nomi-nal feedback current - 06 fC range 59

351 ToPix v3 transfer function for halvedfeedback current - full range 59

352 ToPix v3 transfer function for halvedfeedback current - 06 fC range 59

353 Dead pixel counts in the tested as-semblies 60

354 Source profile obtained from S8 (Epi-75) assembly using a 90Sr source 60

355 Pixel matrix layout 60

356 Scheme of the assembly 61

357 Photograph of epitaxial sensor pro-duced at FBK 61

358 S2 module two readout chips arebump bonded to a single sensor 61

359 Module configuration of a small disk 61

360 Module configuration of a big disk 62

361 The complete forward pixel assemblyscheme 62

362 Busses for different pixel modules 62

363 Scheme of the SU8 technique 63

364 SU8 strips 63

365 Slope of an SU8 strip 63

366 Staircase multilayer 63

367 Bus with module controllers 64

41 Floorplan of wafer with full size pro-totype sensors 67

42 Wafer including Prototype Sensors 68

43 Layout detail of MVD-Barrel DSSDSensors 68

44 I-V-Curves of Si-Strip prototypeSensors 69

45 Capacitance characteristics of largeprototype Sensor 69

46 Effective bulk doping concentrationvs 1 MeV equiv neutron fluence 69

47 Annealing behaviour of leakage cur-rent characteristics 70

48 Annealing behaviour of capacitancecharacteristics 70

49 Dimensions of the strip wedge sensor 70

410 Illustration of the pad geometry atthe top edge of the sensor 70

171

411 Schematic of modified ToPix ASICfor strip readout 73

412 Schematic of the Module Data Con-centrator 74

413 Basic concept for the hybridisationof double-sided strip detectors 75

414 Illustration of the hybridisation con-cept for the strip barrel part 76

415 Schematic cross section of a strip hy-brid structure in the barrel part 77

416 Illustration of the hybridisation con-cept for the strip disks 77

417 Layout of pitch adaptor hybrid 77

418 Photograph of a finally assembleddetector module for the laboratorytest setup 78

419 Photograph of the fabricated pitchadapter and schematics of the padgeometry 78

420 Results obtained with a laboratorysetup for double-sided strip detectors 79

51 GBT and VL scheme 82

52 GBT data packet format 82

53 Architecture of the lower levels of thePANDA DAQ system 83

54 Main components of the MMB 83

55 Low mass cables implementing alu-minium microstrips 86

56 Linear resistance measured for thefirst 20 tracks of each sample 86

57 Total jitter evaluated for just the mi-crostrips without any transceivers 87

58 Example of an eye diagram 87

59 Longitudinal cross-section of theMVD and the cross-pipe sector 88

510 Maximum vertical displacement ofthe frame under static load 88

511 Cams system Cylindrical cam onupper end for fixing the half frame 89

512 Cams system The V insert 89

513 Cams system The U insert 89

514 Structure of a pixel barrel supermodule 90

515 Transverse section of a pixel barrelsuper module 90

516 Support structure of the pixel barrel 90

517 Half barrel assembled on the conesupport 91

518 The cut of the shape of the disk fromthe plate with wire EDM 91

519 Machining of the disks elements 91

520 Sketch of the set of half disks com-posing half of the pixel forward de-tector 92

521 Sketch of the support structure forthe strip disks 92

522 Sketch of the common support barrelfor BL3 and BL4 93

523 Sketch of the set of two strip supermodules with a common cooling pipe 93

524 The Pixel volume with 2 barrels and6 disks 95

525 Test results Youngrsquos modulus [MPa]vs radiation levels for POCO FOAM 95

526 Test results Youngrsquos modulus [Mpa]vs radiation levels for POCO HTC 95

527 Test results thermal conductivity[W(m middotK)] vs radiation levels forPOCO FOAM 95

528 Test results thermal conductivity[W(mmiddotK)] vs radiation levels forPOCO HTC 95

529 Half small disk in carbon foam com-pleted of chips detectors coolingtube and fitting 96

530 Half big disk in carbon foam com-pleted of chips detectors coolingtube and plastic fittings 96

531 FEM analysis results with real con-figuration 97

532 FEM analysis results with test con-figuration 97

533 Carbon foam disk with 6 insertedcooling pipes and 54 resistors(dummy chips) 97

534 Test results Thermal map of diskprototype dissipating 94 W 97

535 Pressure drop [bar] in a S-shape tubevs mass flow rate [lmin] 97

536 Pressure drop [bar] in a U-shape tubevs mass flow rate [lmin] 98

172

537 Prototype of two staves of the secondlayer of the barrel with dummy chips(resistors) 98

538 Test result Thermal map of twostaves prototype 98

539 Schematic view of the cooling con-cept of the strip barrel part 99

540 FEM analysis result for one barrelsection of the strip part 99

541 FEM analysis result for one disk sec-tion of the strip part 99

542 Simple hydraulic scheme of the cool-ing plant 100

543 Preliminary draft of the cooling plant 100

61 Schematic chain of processing 103

62 Basic approach for the introductionof CAD models into physics simula-tions 104

63 Overall view and main structure ofthe MVD model as introduced intophysics simulations 105

64 Schematic sensor profile view andthe linear digitization model with acharge cloud 106

65 Simple model of the signal shape atthe amplifier output 106

66 Illustration of the probability for apixel or strip to produce a noise hit(grey area) 107

67 Riemann efficiency for different par-ticle momenta 108


69 dEdx in the MVD as a function ofthe momentum 110

610 Separation power as a function of themomentum 111

611 Fit function of the mean values of themedian distribution as a function ofthe momentum for different particlespecies 111

612 Efficiency and contamination forprotons and kaons 112

613 Damage factor for different particletypes and momenta 113

614 Dose distribution in the strip part 113

615 Dose distribution in the pixel part 113

616 Dose distribution for Hydrogen andXenon targets 114

617 Dose distribution for various targetsand momenta 114

618 Detector coverage for pions with amomentum of 1 GeVc and respec-tive contributions of the pixel andthe barrel part 115

619 Comparison of the detector coveragefor low-momentum particles 115

620 Yield of different hit counts in fourdifferent detector regions 116

621 Distances of the first MVD hit pointto the nominal IP 117

622 2D material map and extracted 1Dprofile for the full MVD 118

623 Determination of hot spots in thematerial map 118

624 Maps of the contributing materialbudget in terms of the fractional ra-diation length (XX0) 120

625 Contributions of different functionaldetector parts to the overall materialbudget 120

626 Extracted average count rates forantiproton-proton reactions ob-tained in the different detectorlayers 122

627 Resolution on pixel and strip sensorsfor 1 GeVc pions 125

628 Resolutions obtained with a fixed θand an uniform φ distribution 125

629 Vertexing results obtained placingdefined vertices in different pointsalong the longitudinal axis of the MVD126

630 Scan of the vertexing behavior mov-ing vertices on a circle of radius 1 mmlaying in the transverse plane 126

631 Mapping of the vertexing resultschanging the distance from the nom-inal interaction point 126

632 Momentum resolution values ob-tained for π+ and πminus 126

633 Distributions of the Jψ vertices re-constructed from their decay in thenominal interaction point 128

634 Mass distribution for Jψ and ψ(2S)candidates 128

173

635 Distribution of the Jψ mass ob-tained with the missing mass methodknowing the initial ψ(2S) state 128

636 Pion polar angle and momentum dis-tributions from MC-truth and fromthe reconstruction 129

637 Electron-positron polar angle andmomentum distributions from MC-truth and from the reconstruction 129

638 Reconstructed Jψ vertices 129

639 Jψ Reconstructed mass and missingmass distributions 130

640 Coordinates of the reconstructedψ(2S) vertices 130

641 Mass distribution for the ψ(2S) can-didates 130

642 D+Dminus max opening angle 131

643 D+ and Dminus decay lengths 131

644 D+ and Dminus mass resolutions 132

645 D0 and D0

decay lengths 133

646 D0 and D0

mass resolutions 133

647 6-prong background cross sections 133

71 Generalised timeline of the PANDAMVD project 141

72 Work packages of the PANDA MVDproject 142

A1 The Bonn tracking station set up atthe COSY accelerator (Julich) 145

A2 Layout of the DAQ chain 146

A3 Structure of the analysis tools 146

A4 Energy loss measured in one sensorwith beams of protons of two differ-ent momenta 146

A5 Top view of the setup used forthe measurements performed rotat-ing one double sided sensor 146

A6 Peak energy loss of 4 GeV electronsmeasured in one sensor as a functionof the rotation angle of the modulewith respect to the beam direction 147

A7 Simulation of the measurementsshown in figure A6 147

A8 Measured cluster size (in terms ofdigis) as a function of the rotationangle 147

A9 Simulation of the effect of rotationon the cluster size 147

A10 Use of the second box to scan differ-ent longitudinal positions 148

A11 Results of the longitudinal scan per-formed at DESY with 3 GeV electrons148

A12 Setup used to measure the multiplein different carbon volumes 149

A13 Distributions of the projectedscattering angles obtained with295 GeVc protons and the previ-ously described scattering volumes 149

B1 Schematic drawing of the JulichReadout System 151

B2 Schematic overview of the JulichReadout System 152

B3 The digital readout board A of theReadout System 153

B4 The digital readout board B ML605from Xilinx 153

B5 Schematic firmware setup 154

C1 Schematic side view of the barrel sen-sors covering the φ regions aroundthe target pipe 157

C2 Configurations used for the simulations157

D1 The basic alignment problem thecorrect module geometry must be de-duced by studying the deviation ofthe hit positions with respect to thetracks 159

D2 Schematic view of the four basictypes of linear transformations 161

174

175

List of Tables

11 Experimental requirements and op-eration modes of HESR 5

12 Maximum achievable cycle averagedluminosity for different H2 target se-tups 8

13 Expected luminosities for heavier nu-clear targets at PANDA 9

14 Estimate on the expected event ratesat PANDA 10

21 Strange and charmed candidates foridentification by means of their de-layed decay 24

22 Selection of benchmark channels re-quiring an optimum performance ofthe vertex tracking 24

23 Minimum momentum |~pmin| for dif-ferent particle species needed to tra-verse the beryllium pipe 25

24 Compilation of design parameters forthe MVD 31

25 Positions of the active sensor vol-umes within the different detectorlayers 31

31 Thickness and resistivity of the epi-taxial layers of the first assemblies 36

32 Thickness and resistivity featuringthe epitaxial layer of the wafers usedfor the radiation damage tests 37

33 Pre irradiation full depletion voltagevalues as measured for epitaxial diodes 37

34 Sensor numbers 41

35 Epitaxial wafer properties for sensorproduction 42

36 Specification summary for the front-end ASIC 42

37 ToPix v1 electrical tests result 48

38 Ions used in the beam test 51

39 Fit parameters of the shift registerand the 12b counter in standard elec-tronics 52

310 SEU cross section in the hadron en-vironment corrected for the under-estimation of the SV 52

311 SEU (bitminus1 sminus1) in the ToPix v2 pro-totype 52

312 SEU (chipminus1 hminus1) for the PANDA en-vironment in the final ToPix prototype 52

41 Specifications of the prototype stripsensors 68

42 Design parameters of the wedge sensor 71

43 Requirements for the strip front-endASIC 72

44 Comparison of different front-endsfor strip detectors 74

45 Specifications of the hybrid carriermaterial 76

51 M55JLTM110-5 Composite properties 88

52 POCO HTC properties 94

53 POCO FOAM properties 94

61 Main setups for the studies of the de-tector coverage 114

62 Relative detector coverage in the ac-tive detector region obtained with pi-ons of both charge conjugations 117

63 Main results of the count rate studyfor antiproton-proton reactions 123

64 Simulation parameters 124

65 D meson counts 132

66 D+ and Dminus vertex resolutions 132

67 D0 and D0

vertex resolutions 132

A1 The position of the four boxes and ofthe ideal sensor 148

A2 Results of the simulations performedwith different setups 149

A3 Comparison between the sigma ofGaussian fits on the distributions ofthe projected scattering angles 150

C1 Results obtained with an artificialrotation of track candidates beforethe vertex finding (POCA) 157

C2 Resolutions obtained with the differ-ent setups and kinematics 158

Preface

1 The Panda Experiment and its Tracking Concept

11 The Panda Experiment


112 High Energy Storage Ring ndash HESR

113 Targets


12 The Panda Detector




124 Infrastructure

13 The Charged Particle Tracking System

131 Basic Approach



21 General Overview


23 Basic Detector Requirements

24 MVD Layout




31 Hybrid Pixel Assembly Concept

32 Sensor




324 Technology Choice Epi vs Oxygen

325 Production


331 Requirements


333 Front-End ASIC


335 ASIC Prototypes

34 Hybridisation

341 Bump Bonding

35 Single Chip Assembly Prototype

36 Module

37 Bus-Flex Hybrid


41 Double-Sided Silicon Strip Detectors (DSSD)

411 Barrel Sensors

412 Wedge Sensors


421 Requirements

422 Options

43 Module Data Concentrator ASIC

431 Architecture

432 Implemented Feature Extraction Algorithms


44 Hybridisation

441 Overview




445 Test Assembly

5 Infrastructure





531 Introduction

532 Powering Concept for the Pixel Part

533 Powering Concept for the Strip Part

54 Cables

541 Requirements

542 Signal Cable

543 Power Cable




553 Support Structures of the Strip Part




563 Cooling Plant

57 DCS



62 Detector Model of the MVD



632 Digitization

633 Noise Emulation



641 MVD Tracklet Finding and Fitting

642 Track Finding

643 Track Fitting




652 Statistical Approach for the Energy Loss Parametrisation




663 Material Budget

664 Rate Estimation

67 Resolution and Performance Studies

671 Hit Resolution



681 Detector Setup

682 Benchmark Channel pp (2S) J+ -


684 Considerations About the Vertexing


71 Quality Control and Assembly

711 Pixel

712 Strips

713 Integration

72 Safety

73 Timeline and Work Packages

Appendices

A The Bonn Tracking Station as a Validation for the Simulation Framework








B Juumllich Readout System

B1 Overview

B2 Basic Concepts



B23 Firmware


B3 Conclusion


D Alignment

D1 Introduction

D2 General Strategies and Techniques for VertexTracking Detectors

D21 Software Alignment of Tracking Modules and Modern Tools

List of Acronyms

List of Figures

List of Tables





PANDA Collaboration





PANDA Collaboration

30th November 2011
































G Stancari


























T Zerguerras













R Wheadon L Zotti

















MVD Group















R Jakel2 D-L Pohl4

V Jha4 T Quagli9



Preface




ix

Contents

Preface vii





113 Targets 6














24 MVD Layout 27





32 Sensor 36





325 Production 41


331 Requirements 42





34 Hybridisation 59

341 Bump Bonding 59


36 Module 61







421 Requirements 71

422 Options 71


431 Architecture 74



44 Hybridisation 75

441 Overview 75





5 Infrastructure 81





531 Introduction 83



54 Cables 85

541 Requirements 85

542 Signal Cable 85

543 Power Cable 87










57 DCS 101
































711 Pixel 137

712 Strips 138

713 Integration 138

72 Safety 139


Appendices 143










B1 Overview 151




B23 Firmware 153


B3 Conclusion 154


D Alignment 159

D1 Introduction 159




List of Figures 169

List of Tables 175

1











































Operation Modes




15 to 89 GeVc




15 to 15 GeVc









113 Targets



Cluster Jet Target























Cluster jet target







Pellet target












texpτ

]texp + tprep

(12)













p the



(RArp

+ 1

)2

(13)


Event Rates




































Beam-Target System


Solenoid Magnet























Barrel DIRC











Muon Detectors














Dipole Magnet


Forward Trackers









radicE [30] has

been achieved




Luminosity Monitor





124 Infrastructure






131 Basic Approach








Acceptance






























Material Budget


References




































23


21 General Overview























eg K0SK



K0Sπ







Σ+ 8018(26) ps 2404 cm pπ0 ((5157plusmn 030))

nπ+ ((4831plusmn 030))







0((23plusmn 06)














































24 MVD Layout
























Sensor Geometry














Number of sensors




60 (6 chips size)

Total 176 254




Total 810 1324



MainSub-layer



BarrelInner ring

2522 2180 2275


BarrelInner ring

50475 4730 4785




DiskSdk 1 front

lt 50 - 1170 3656 20 22197 199


DiskSdk 1 front

lt 50 - 1170 3656 40 42397 399


DiskLdk 1 front

lt 95 - 1170 7396 70 72697 699


DiskLdk 2 front

lt 95 - 1170 7396 100 102997 999


Disk

Ldk 3 front- - 1170 7396

160

1501477 1479

layer 5

Ldk 3 rear 1521 1523

StripDk 1 L- - 7433 13115 1625

1600 1603

StripDk 1 S 1650 1653

Disk

Ldk 4 front- - 1170 7396

230

2202177 2179

layer 6

Ldk 4 rear 2221 2223

StripDk 2 S- - 7433 13115 2075

2047 2050

StripDk 2 L 2097 2100
















References








35










32 Sensor













Results
























Epi-50 HR 435 006

Epi-75 HR 56 01

Epi-100 HR 59 02

Epi-50 MR 491 007

Epi-75 MR 82 03

Epi-100 MR 104 06

Epi-75 LR 426 13
























Tests















325 Production



S2 S4 S5 S6 Total

Barrel1 6 8 0 0 14

Barrel2 0 0 6 44 50

Disk1 6 2 0 0 8

Disk2 6 2 0 0 8

Disk3 4 4 12 4 24

Disk4 4 4 12 4 24

Disk5 4 4 12 4 24

Disk6 4 4 12 4 24

Total 34 28 54 60 176

Chip Number 68 112 270 360 810







Cz substrate




Dopant Sb



Epitaxial Layer

Conductivity type n

Dopant P





331 Requirements


























333 Front-End ASIC



Pixel Cell










ToPix Architecture











335 ASIC Prototypes


ToPix Version 1


















ToPix Version 2









































SEU Study






σSEU =Nerrors

Φ middotNbit(31)


σ = σ0

[1minus eminus

(EdepminusE0

W

)s](32)



bull Eth=04688 MeV


bull W = 4375 MeV











Σ [cm2bit]

CR SR CNT










pminus p 3 130 207

pminus N 1 27 42




ToPix Version 3
























































34 Hybridisation


341 Bump Bonding
























36 Module
















37 Bus-Flex Hybrid























References





























67





411 Barrel Sensors








General





guard rings 8


S1

n-side strips 896

p-side strips 512

pitch 65 microm


S2

n-side strips 512

p-side strips 512

pitch 65 microm


S3

n-side strips 384

p-side strips 384

pitch 50 microm


S4

n-side strips 128

p-side strips 128

pitch 65 microm










0

5

10

15

20

0 50 100 150 200 250 300 350

Leak

age

Cur

rent

microA

Bias Voltage V


01020406070809101112131415

161718192022232425262728








Bias Voltage [V]0 50 100 150 200

Cap

acit

ance

[p

F]

0

5

10

15

20

25

30

35

whole Sensor

strip 21

strip 27

strip 524

strip 736

strip 740






Bias Voltage [V]0 20 40 60 80 100 120

Lea

kag

e C

urr

ent

[uA

cm

^3]

1

10

210

310

before irradiation

20h

310h

5100h


Bias Voltage [V]0 20 40 60 80 100 120

Cap

acit

ance

[p

F]

3


20h

310h

5100h



412 Wedge Sensors









Stereo angle 15


Total area 1688 mm2

(Active area) (899)



421 Requirements



422 Options








width le 8 mm

depth le 8 mm




Input Compliance

sensor capacitances








Signal





Power


Dynamical








40000 sminus1 pAu


30000 sminus1 pAu

Interface







-+

-+

-+

-+


Current buer

Cf

Cz

Rz

Rf

Cd-

+

-+

-+

-+

-+


Cf

Cz

Rz

Rf

Cd-

+Current buer





STS-XYTER


FSSR2






















431 Architecture







48frasl

frasl

frasl

48

48








44 Hybridisation

441 Overview



















conductor type Cu




















445 Test Assembly










References



















81

5 Infrastructure



























531 Introduction








Power Regulators










54 Cables


541 Requirements



542 Signal Cable




Al Prototypes



Folded Layout




Straight Layout



Results











543 Power Cable





bull pixel barrels

bull pixel disks

bull strip barrels

bull strip disks












CTE resin (ppmC) 60
















Pixel Barrels












Pixel Disks















Strip Barrels



Strip Disks














POCO HTC

Density 09 gcm3





POCO FOAM

Density 055 gcm3
















Disk Cooling System











































563 Cooling Plant











57 DCS





bull coolant flow







References



















103











































632 Digitization



x

Q

chargecloud

particle track

sensor





Qi =Q



Qij =Q







(xk minus xradic

2σC

)+

radic2σ2

C


2σ2C









Q0 1000 2000 3000 4000 5000 6000

-710

-610

-510

-410

-310

-210

-110

1Threshold

NoiseDistribution



633 Noise Emulation



2erfc

(Qthrradic

2 middot σnoise

)(64)




Cluster Finding


Hit Reconstruction



radic12



x =

sumi qixisumi qi

(65)


radic12 which will








tTW =QThr[e

minus]











642 Track Finding




643 Track Fitting




POCA Finder




~V =

sumni=0j=i+1D

minus1ij

~Vijsumni=0j=i+1D

minus1ij

(67)


D =

nsumi=0j=i+1

Dminus1ij

minus1

(68)



















∆E∆x =


(69)








Mk(x1 x2 xn) =

(1

n

nsumi=1

xki

) 1k

(610)




σp12 + σp22 (611)





dE

dx=

a

p2


](612)













10-10 10-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 100 101 102 103 104

E[MeV]

10-5

10-4

10-3

10-2

10-1

100

101

102

103

104

D(E

)95

MeV

mb


neutrons


protons

pions Huhtinen

pions

electrons Summers

electrons






t =A

R=

A

σ middot L(613)












0

50

100

150

200

250

300

350

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

10

9N

EQ(

mm

2 middot1

07 s

)

p [GeVc]

Xe

Xe_max

N

N_max

p

p_max








































1 2 3 4 5 6 Total

015 GeVc+ 63 76 223 385 169 36 950

minus 45 90 210 384 171 34 932

05 GeVc+ 61 64 230 397 169 38 949

minus 54 71 224 409 166 36 945

10 GeVc+ 60 64 227 400 171 37 952

minus 56 69 224 408 166 38 950

15 GeVc+ 60 64 226 402 171 38 952

minus 57 68 224 408 166 39 952




663 Material Budget


XX0 =sumj

ρj middot LjX0j

(614)











〈XX0〉+ 2 middot 〈σθi〉 = 91 (615)




























664 Rate Estimation















Pixel part









Strip part
























671 Hit Resolution













Mean 109e-05

RMS 0002158

[cm]reco-xmcx-001 -0005 0 0005 0010

200400600800

100012001400


Mean 109e-05


Mean 1127e-05

RMS 0002886

[cm]reco-xmcx-001 -0005 0 0005 0010

200

400

600

800

1000


Mean 1127e-05

RMS 0002886 =1clust

Strip Resolution n


Mean -3275e-06

RMS 0001379

ndf 2

χ 4465 18



Sigma 00000074plusmn 00006904

[cm]reco-xmcx-001 -0005 0 0005 0010

500

1000

1500

2000

2500diffnxmul

Entries 51326

Mean -3275e-06

RMS 0001379

ndf 2

χ 4465 18





diffnymul

[cm]reco-xmcx-001 -0005 0 0005 0010

500

1000

1500

2000

2500

diffnymul gt1clust

Strip Resolution n

















681 Detector Setup





bull GEM 3 stations





bull Disc DIRC

bull Barrel DIRC

bull EMC crystals











Jψ rarr e+eminus

























m]microct [

0 500 1000 1500 2000

1

10

210

310







comb nonres





Vertex r and z cut 15004 689 1060 17255 194 543

4C kinematic fit 1058 25 0 1592 18 2




D+ Poca PRG KinVtx









]2M[GeVc17 18 19 2 21

210

310

Reco

Vtx-Cut

Vtx-Cut amp 4C-Fit


2 =87 MeVc4Cσ





D0 Poca PRG KinVtx



σz microm 884 949 902

D0

Poca PRG KinVtx



σz microm 884 941 893

Table 67 D0 and D0


6832 pprarr D0D0







m]microct [

0 200 400 600 800 1000 1200 1400

1

10

210

310




]2M[GeVc17 18 19 2 21

210

310

Reco

Vtx-Cut

Vtx amp 4C-Fit


2 =58 MeVc4Cσ




SB asymp 003 (617)







[GeVc] lab

p0 2 4 6 8 10 12 14 16 18

[m

b]

-210

-110

1

10

210

310

p totalp

6 prong

-3+3

0-3+3

-2+2K2





115





References

















































137





711 Pixel



















712 Strips




713 Integration





72 Safety












Acknowledgments





Task NamePanda MVD






CoolingDCS


Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q12009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019





143

Appendices

144

145

















bull FIFO buffering







146










147








1 α = arctan(

pitchthickness

)= arctan

(50microm300microm

)= 946

148










B1 B2 Device B3 B4


A 16 86 110 145 1855

B 90 100 110 120 130

C 65 85 110 139 159



149


A 56 53

B 16 16

C 34 34















150


GeVc mrad mrad

1 cm C p+ 295 102 101

2 cm C p+ 295 134 133


Air eminus 30 0423 0476

Air eminus 54 0243 0284







References



151


B1 Overview







bull FPGA Firmware


bull Readout PC




B2 Basic Concepts


Modularity






152























153







bull 16 LVDS inputs












bull LC-Display



B23 Firmware





154
















B3 Conclusion


References



155







156

157




x 883 694

y 694 883






158









159

D Alignment

D1 Introduction












160

















χ2 =sum

events

sumtracks

sumhits

∆2i σ

2i (D1)


161



xfit =

nsumi=1

aidi +

νsumj=1

αjδj (D2)



162









163







References








html







164

165

3D 3-Dimensional









BTeV B meson TeV









CNT Counter






CT Central Tracker

Cz Czochralski




DC Direct Current















FE Front-End










FZ Floating Zone




GBT GigaBit



166



GR Guard Ring










HV High Voltage





LD Laser Driver









MC Monte-Carlo



















PD Photo Diode




PMT Photomultiplier





RF Radio Frequency







167





SP Separation Power

SR Shift Register






TOF Time-of-Flight

ToPix Torino Pixel


TRx Transceiver


UI Unit Interval






VL Versatile Link





168

169

List of Figures












































170



322 ToPix v2 die 49










































364 SU8 strips 63














171















































172










































173











645 D0 and D0

decay lengths 133

646 D0 and D0



























174

175

List of Tables




































67 D0 and D0







Preface





113 Targets






124 Infrastructure


131 Basic Approach



21 General Overview



24 MVD Layout





32 Sensor





325 Production


331 Requirements


333 Front-End ASIC


335 ASIC Prototypes

34 Hybridisation

341 Bump Bonding


36 Module

37 Bus-Flex Hybrid



411 Barrel Sensors

412 Wedge Sensors


421 Requirements

422 Options


431 Architecture



44 Hybridisation

441 Overview




445 Test Assembly

5 Infrastructure





531 Introduction



54 Cables

541 Requirements

542 Signal Cable

543 Power Cable








563 Cooling Plant

57 DCS






632 Digitization

633 Noise Emulation




642 Track Finding

643 Track Fitting








663 Material Budget

664 Rate Estimation


671 Hit Resolution



681 Detector Setup






711 Pixel

712 Strips

713 Integration

72 Safety


Appendices










B1 Overview

B2 Basic Concepts



B23 Firmware


B3 Conclusion


D Alignment

D1 Introduction



List of Acronyms

List of Figures

List of Tables





PANDA Collaboration

30th November 2011
































G Stancari


























T Zerguerras













R Wheadon L Zotti

















MVD Group















R Jakel2 D-L Pohl4

V Jha4 T Quagli9



Preface




ix

Contents

Preface vii





113 Targets 6














24 MVD Layout 27





32 Sensor 36





325 Production 41


331 Requirements 42





34 Hybridisation 59

341 Bump Bonding 59


36 Module 61







421 Requirements 71

422 Options 71


431 Architecture 74



44 Hybridisation 75

441 Overview 75





5 Infrastructure 81





531 Introduction 83



54 Cables 85

541 Requirements 85

542 Signal Cable 85

543 Power Cable 87










57 DCS 101
































711 Pixel 137

712 Strips 138

713 Integration 138

72 Safety 139


Appendices 143










B1 Overview 151




B23 Firmware 153


B3 Conclusion 154


D Alignment 159

D1 Introduction 159




List of Figures 169

List of Tables 175

1











































Operation Modes




15 to 89 GeVc




15 to 15 GeVc









113 Targets



Cluster Jet Target























Cluster jet target







Pellet target












texpτ

]texp + tprep

(12)













p the



(RArp

+ 1

)2

(13)


Event Rates




































Beam-Target System


Solenoid Magnet























Barrel DIRC











Muon Detectors














Dipole Magnet


Forward Trackers









radicE [30] has

been achieved




Luminosity Monitor





124 Infrastructure






131 Basic Approach








Acceptance






























Material Budget


References




































23


21 General Overview























eg K0SK



K0Sπ







Σ+ 8018(26) ps 2404 cm pπ0 ((5157plusmn 030))

nπ+ ((4831plusmn 030))







0((23plusmn 06)














































24 MVD Layout
























Sensor Geometry














Number of sensors




60 (6 chips size)

Total 176 254




Total 810 1324



MainSub-layer



BarrelInner ring

2522 2180 2275


BarrelInner ring

50475 4730 4785




DiskSdk 1 front

lt 50 - 1170 3656 20 22197 199


DiskSdk 1 front

lt 50 - 1170 3656 40 42397 399


DiskLdk 1 front

lt 95 - 1170 7396 70 72697 699


DiskLdk 2 front

lt 95 - 1170 7396 100 102997 999


Disk

Ldk 3 front- - 1170 7396

160

1501477 1479

layer 5

Ldk 3 rear 1521 1523

StripDk 1 L- - 7433 13115 1625

1600 1603

StripDk 1 S 1650 1653

Disk

Ldk 4 front- - 1170 7396

230

2202177 2179

layer 6

Ldk 4 rear 2221 2223

StripDk 2 S- - 7433 13115 2075

2047 2050

StripDk 2 L 2097 2100
















References








35










32 Sensor













Results
























Epi-50 HR 435 006

Epi-75 HR 56 01

Epi-100 HR 59 02

Epi-50 MR 491 007

Epi-75 MR 82 03

Epi-100 MR 104 06

Epi-75 LR 426 13
























Tests















325 Production



S2 S4 S5 S6 Total

Barrel1 6 8 0 0 14

Barrel2 0 0 6 44 50

Disk1 6 2 0 0 8

Disk2 6 2 0 0 8

Disk3 4 4 12 4 24

Disk4 4 4 12 4 24

Disk5 4 4 12 4 24

Disk6 4 4 12 4 24

Total 34 28 54 60 176

Chip Number 68 112 270 360 810







Cz substrate




Dopant Sb



Epitaxial Layer

Conductivity type n

Dopant P





331 Requirements


























333 Front-End ASIC



Pixel Cell










ToPix Architecture











335 ASIC Prototypes


ToPix Version 1


















ToPix Version 2









































SEU Study






σSEU =Nerrors

Φ middotNbit(31)


σ = σ0

[1minus eminus

(EdepminusE0

W

)s](32)



bull Eth=04688 MeV


bull W = 4375 MeV











Σ [cm2bit]

CR SR CNT










pminus p 3 130 207

pminus N 1 27 42




ToPix Version 3
























































34 Hybridisation


341 Bump Bonding
























36 Module
















37 Bus-Flex Hybrid























References





























67





411 Barrel Sensors








General





guard rings 8


S1

n-side strips 896

p-side strips 512

pitch 65 microm


S2

n-side strips 512

p-side strips 512

pitch 65 microm


S3

n-side strips 384

p-side strips 384

pitch 50 microm


S4

n-side strips 128

p-side strips 128

pitch 65 microm










0

5

10

15

20

0 50 100 150 200 250 300 350

Leak

age

Cur

rent

microA

Bias Voltage V


01020406070809101112131415

161718192022232425262728








Bias Voltage [V]0 50 100 150 200

Cap

acit

ance

[p

F]

0

5

10

15

20

25

30

35

whole Sensor

strip 21

strip 27

strip 524

strip 736

strip 740






Bias Voltage [V]0 20 40 60 80 100 120

Lea

kag

e C

urr

ent

[uA

cm

^3]

1

10

210

310

before irradiation

20h

310h

5100h


Bias Voltage [V]0 20 40 60 80 100 120

Cap

acit

ance

[p

F]

3


20h

310h

5100h



412 Wedge Sensors









Stereo angle 15


Total area 1688 mm2

(Active area) (899)



421 Requirements



422 Options








width le 8 mm

depth le 8 mm




Input Compliance

sensor capacitances








Signal





Power


Dynamical








40000 sminus1 pAu


30000 sminus1 pAu

Interface







-+

-+

-+

-+


Current buer

Cf

Cz

Rz

Rf

Cd-

+

-+

-+

-+

-+


Cf

Cz

Rz

Rf

Cd-

+Current buer





STS-XYTER


FSSR2






















431 Architecture







48frasl

frasl

frasl

48

48








44 Hybridisation

441 Overview



















conductor type Cu




















445 Test Assembly










References



















81

5 Infrastructure



























531 Introduction








Power Regulators










54 Cables


541 Requirements



542 Signal Cable




Al Prototypes



Folded Layout




Straight Layout



Results











543 Power Cable





bull pixel barrels

bull pixel disks

bull strip barrels

bull strip disks












CTE resin (ppmC) 60
















Pixel Barrels












Pixel Disks















Strip Barrels



Strip Disks














POCO HTC

Density 09 gcm3





POCO FOAM

Density 055 gcm3
















Disk Cooling System











































563 Cooling Plant











57 DCS





bull coolant flow







References



















103











































632 Digitization



x

Q

chargecloud

particle track

sensor





Qi =Q



Qij =Q







(xk minus xradic

2σC

)+

radic2σ2

C


2σ2C









Q0 1000 2000 3000 4000 5000 6000

-710

-610

-510

-410

-310

-210

-110

1Threshold

NoiseDistribution



633 Noise Emulation



2erfc

(Qthrradic

2 middot σnoise

)(64)




Cluster Finding


Hit Reconstruction



radic12



x =

sumi qixisumi qi

(65)


radic12 which will








tTW =QThr[e

minus]











642 Track Finding




643 Track Fitting




POCA Finder




~V =

sumni=0j=i+1D

minus1ij

~Vijsumni=0j=i+1D

minus1ij

(67)


D =

nsumi=0j=i+1

Dminus1ij

minus1

(68)



















∆E∆x =


(69)








Mk(x1 x2 xn) =

(1

n

nsumi=1

xki

) 1k

(610)




σp12 + σp22 (611)





dE

dx=

a

p2


](612)













10-10 10-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 100 101 102 103 104

E[MeV]

10-5

10-4

10-3

10-2

10-1

100

101

102

103

104

D(E

)95

MeV

mb


neutrons


protons

pions Huhtinen

pions

electrons Summers

electrons






t =A

R=

A

σ middot L(613)












0

50

100

150

200

250

300

350

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

10

9N

EQ(

mm

2 middot1

07 s

)

p [GeVc]

Xe

Xe_max

N

N_max

p

p_max








































1 2 3 4 5 6 Total

015 GeVc+ 63 76 223 385 169 36 950

minus 45 90 210 384 171 34 932

05 GeVc+ 61 64 230 397 169 38 949

minus 54 71 224 409 166 36 945

10 GeVc+ 60 64 227 400 171 37 952

minus 56 69 224 408 166 38 950

15 GeVc+ 60 64 226 402 171 38 952

minus 57 68 224 408 166 39 952




663 Material Budget


XX0 =sumj

ρj middot LjX0j

(614)











〈XX0〉+ 2 middot 〈σθi〉 = 91 (615)




























664 Rate Estimation















Pixel part









Strip part
























671 Hit Resolution













Mean 109e-05

RMS 0002158

[cm]reco-xmcx-001 -0005 0 0005 0010

200400600800

100012001400


Mean 109e-05


Mean 1127e-05

RMS 0002886

[cm]reco-xmcx-001 -0005 0 0005 0010

200

400

600

800

1000


Mean 1127e-05

RMS 0002886 =1clust

Strip Resolution n


Mean -3275e-06

RMS 0001379

ndf 2

χ 4465 18



Sigma 00000074plusmn 00006904

[cm]reco-xmcx-001 -0005 0 0005 0010

500

1000

1500

2000

2500diffnxmul

Entries 51326

Mean -3275e-06

RMS 0001379

ndf 2

χ 4465 18





diffnymul

[cm]reco-xmcx-001 -0005 0 0005 0010

500

1000

1500

2000

2500

diffnymul gt1clust

Strip Resolution n

















681 Detector Setup





bull GEM 3 stations





bull Disc DIRC

bull Barrel DIRC

bull EMC crystals











Jψ rarr e+eminus

























m]microct [

0 500 1000 1500 2000

1

10

210

310







comb nonres





Vertex r and z cut 15004 689 1060 17255 194 543

4C kinematic fit 1058 25 0 1592 18 2




D+ Poca PRG KinVtx









]2M[GeVc17 18 19 2 21

210

310

Reco

Vtx-Cut

Vtx-Cut amp 4C-Fit


2 =87 MeVc4Cσ





D0 Poca PRG KinVtx



σz microm 884 949 902

D0

Poca PRG KinVtx



σz microm 884 941 893

Table 67 D0 and D0


6832 pprarr D0D0







m]microct [

0 200 400 600 800 1000 1200 1400

1

10

210

310




]2M[GeVc17 18 19 2 21

210

310

Reco

Vtx-Cut

Vtx amp 4C-Fit


2 =58 MeVc4Cσ




SB asymp 003 (617)







[GeVc] lab

p0 2 4 6 8 10 12 14 16 18

[m

b]

-210

-110

1

10

210

310

p totalp

6 prong

-3+3

0-3+3

-2+2K2





115





References

















































137





711 Pixel



















712 Strips




713 Integration





72 Safety












Acknowledgments





Task NamePanda MVD






CoolingDCS


Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q12009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019





143

Appendices

144

145

















bull FIFO buffering







146










147








1 α = arctan(

pitchthickness

)= arctan

(50microm300microm

)= 946

148










B1 B2 Device B3 B4


A 16 86 110 145 1855

B 90 100 110 120 130

C 65 85 110 139 159



149


A 56 53

B 16 16

C 34 34















150


GeVc mrad mrad

1 cm C p+ 295 102 101

2 cm C p+ 295 134 133


Air eminus 30 0423 0476

Air eminus 54 0243 0284







References



151


B1 Overview







bull FPGA Firmware


bull Readout PC




B2 Basic Concepts


Modularity






152























153







bull 16 LVDS inputs












bull LC-Display



B23 Firmware





154
















B3 Conclusion


References



155







156

157




x 883 694

y 694 883






158









159

D Alignment

D1 Introduction












160

















χ2 =sum

events

sumtracks

sumhits

∆2i σ

2i (D1)


161



xfit =

nsumi=1

aidi +

νsumj=1

αjδj (D2)



162









163







References








html







164

165

3D 3-Dimensional









BTeV B meson TeV









CNT Counter






CT Central Tracker

Cz Czochralski




DC Direct Current















FE Front-End










FZ Floating Zone




GBT GigaBit



166



GR Guard Ring










HV High Voltage





LD Laser Driver









MC Monte-Carlo



















PD Photo Diode




PMT Photomultiplier





RF Radio Frequency







167





SP Separation Power

SR Shift Register






TOF Time-of-Flight

ToPix Torino Pixel


TRx Transceiver


UI Unit Interval






VL Versatile Link





168

169

List of Figures












































170



322 ToPix v2 die 49










































364 SU8 strips 63














171















































172










































173











645 D0 and D0

decay lengths 133

646 D0 and D0



























174

175

List of Tables




































67 D0 and D0







Preface





113 Targets






124 Infrastructure


131 Basic Approach



21 General Overview



24 MVD Layout





32 Sensor





325 Production


331 Requirements


333 Front-End ASIC


335 ASIC Prototypes

34 Hybridisation

341 Bump Bonding


36 Module

37 Bus-Flex Hybrid



411 Barrel Sensors

412 Wedge Sensors


421 Requirements

422 Options


431 Architecture



44 Hybridisation

441 Overview




445 Test Assembly

5 Infrastructure





531 Introduction



54 Cables

541 Requirements

542 Signal Cable

543 Power Cable








563 Cooling Plant

57 DCS






632 Digitization

633 Noise Emulation




642 Track Finding

643 Track Fitting








663 Material Budget

664 Rate Estimation


671 Hit Resolution



681 Detector Setup






711 Pixel

712 Strips

713 Integration

72 Safety


Appendices










B1 Overview

B2 Basic Concepts



B23 Firmware


B3 Conclusion


D Alignment

D1 Introduction



List of Acronyms

List of Figures

List of Tables
































G Stancari


























T Zerguerras













R Wheadon L Zotti

















MVD Group















R Jakel2 D-L Pohl4

V Jha4 T Quagli9



Preface




ix

Contents

Preface vii





113 Targets 6














24 MVD Layout 27





32 Sensor 36





325 Production 41


331 Requirements 42





34 Hybridisation 59

341 Bump Bonding 59


36 Module 61







421 Requirements 71

422 Options 71


431 Architecture 74



44 Hybridisation 75

441 Overview 75





5 Infrastructure 81





531 Introduction 83



54 Cables 85

541 Requirements 85

542 Signal Cable 85

543 Power Cable 87










57 DCS 101
































711 Pixel 137

712 Strips 138

713 Integration 138

72 Safety 139


Appendices 143










B1 Overview 151




B23 Firmware 153


B3 Conclusion 154


D Alignment 159

D1 Introduction 159




List of Figures 169

List of Tables 175

1











































Operation Modes




15 to 89 GeVc




15 to 15 GeVc









113 Targets



Cluster Jet Target























Cluster jet target







Pellet target












texpτ

]texp + tprep

(12)













p the



(RArp

+ 1

)2

(13)


Event Rates




































Beam-Target System


Solenoid Magnet























Barrel DIRC











Muon Detectors














Dipole Magnet


Forward Trackers









radicE [30] has

been achieved




Luminosity Monitor





124 Infrastructure






131 Basic Approach








Acceptance






























Material Budget


References




































23


21 General Overview























eg K0SK



K0Sπ







Σ+ 8018(26) ps 2404 cm pπ0 ((5157plusmn 030))

nπ+ ((4831plusmn 030))







0((23plusmn 06)














































24 MVD Layout
























Sensor Geometry














Number of sensors




60 (6 chips size)

Total 176 254




Total 810 1324



MainSub-layer



BarrelInner ring

2522 2180 2275


BarrelInner ring

50475 4730 4785




DiskSdk 1 front

lt 50 - 1170 3656 20 22197 199


DiskSdk 1 front

lt 50 - 1170 3656 40 42397 399


DiskLdk 1 front

lt 95 - 1170 7396 70 72697 699


DiskLdk 2 front

lt 95 - 1170 7396 100 102997 999


Disk

Ldk 3 front- - 1170 7396

160

1501477 1479

layer 5

Ldk 3 rear 1521 1523

StripDk 1 L- - 7433 13115 1625

1600 1603

StripDk 1 S 1650 1653

Disk

Ldk 4 front- - 1170 7396

230

2202177 2179

layer 6

Ldk 4 rear 2221 2223

StripDk 2 S- - 7433 13115 2075

2047 2050

StripDk 2 L 2097 2100
















References








35










32 Sensor













Results
























Epi-50 HR 435 006

Epi-75 HR 56 01

Epi-100 HR 59 02

Epi-50 MR 491 007

Epi-75 MR 82 03

Epi-100 MR 104 06

Epi-75 LR 426 13
























Tests















325 Production



S2 S4 S5 S6 Total

Barrel1 6 8 0 0 14

Barrel2 0 0 6 44 50

Disk1 6 2 0 0 8

Disk2 6 2 0 0 8

Disk3 4 4 12 4 24

Disk4 4 4 12 4 24

Disk5 4 4 12 4 24

Disk6 4 4 12 4 24

Total 34 28 54 60 176

Chip Number 68 112 270 360 810







Cz substrate




Dopant Sb



Epitaxial Layer

Conductivity type n

Dopant P





331 Requirements


























333 Front-End ASIC



Pixel Cell










ToPix Architecture











335 ASIC Prototypes


ToPix Version 1


















ToPix Version 2









































SEU Study






σSEU =Nerrors

Φ middotNbit(31)


σ = σ0

[1minus eminus

(EdepminusE0

W

)s](32)



bull Eth=04688 MeV


bull W = 4375 MeV











Σ [cm2bit]

CR SR CNT










pminus p 3 130 207

pminus N 1 27 42




ToPix Version 3
























































34 Hybridisation


341 Bump Bonding
























36 Module
















37 Bus-Flex Hybrid























References





























67





411 Barrel Sensors








General





guard rings 8


S1

n-side strips 896

p-side strips 512

pitch 65 microm


S2

n-side strips 512

p-side strips 512

pitch 65 microm


S3

n-side strips 384

p-side strips 384

pitch 50 microm


S4

n-side strips 128

p-side strips 128

pitch 65 microm










0

5

10

15

20

0 50 100 150 200 250 300 350

Leak

age

Cur

rent

microA

Bias Voltage V


01020406070809101112131415

161718192022232425262728








Bias Voltage [V]0 50 100 150 200

Cap

acit

ance

[p

F]

0

5

10

15

20

25

30

35

whole Sensor

strip 21

strip 27

strip 524

strip 736

strip 740






Bias Voltage [V]0 20 40 60 80 100 120

Lea

kag

e C

urr

ent

[uA

cm

^3]

1

10

210

310

before irradiation

20h

310h

5100h


Bias Voltage [V]0 20 40 60 80 100 120

Cap

acit

ance

[p

F]

3


20h

310h

5100h



412 Wedge Sensors









Stereo angle 15


Total area 1688 mm2

(Active area) (899)



421 Requirements



422 Options








width le 8 mm

depth le 8 mm




Input Compliance

sensor capacitances








Signal





Power


Dynamical








40000 sminus1 pAu


30000 sminus1 pAu

Interface







-+

-+

-+

-+


Current buer

Cf

Cz

Rz

Rf

Cd-

+

-+

-+

-+

-+


Cf

Cz

Rz

Rf

Cd-

+Current buer





STS-XYTER


FSSR2






















431 Architecture







48frasl

frasl

frasl

48

48








44 Hybridisation

441 Overview



















conductor type Cu




















445 Test Assembly










References



















81

5 Infrastructure



























531 Introduction








Power Regulators










54 Cables


541 Requirements



542 Signal Cable




Al Prototypes



Folded Layout




Straight Layout



Results











543 Power Cable





bull pixel barrels

bull pixel disks

bull strip barrels

bull strip disks












CTE resin (ppmC) 60
















Pixel Barrels












Pixel Disks















Strip Barrels



Strip Disks














POCO HTC

Density 09 gcm3





POCO FOAM

Density 055 gcm3
















Disk Cooling System











































563 Cooling Plant











57 DCS





bull coolant flow







References



















103











































632 Digitization



x

Q

chargecloud

particle track

sensor





Qi =Q



Qij =Q







(xk minus xradic

2σC

)+

radic2σ2

C


2σ2C









Q0 1000 2000 3000 4000 5000 6000

-710

-610

-510

-410

-310

-210

-110

1Threshold

NoiseDistribution



633 Noise Emulation



2erfc

(Qthrradic

2 middot σnoise

)(64)




Cluster Finding


Hit Reconstruction



radic12



x =

sumi qixisumi qi

(65)


radic12 which will








tTW =QThr[e

minus]











642 Track Finding




643 Track Fitting




POCA Finder




~V =

sumni=0j=i+1D

minus1ij

~Vijsumni=0j=i+1D

minus1ij

(67)


D =

nsumi=0j=i+1

Dminus1ij

minus1

(68)



















∆E∆x =


(69)








Mk(x1 x2 xn) =

(1

n

nsumi=1

xki

) 1k

(610)




σp12 + σp22 (611)





dE

dx=

a

p2


](612)













10-10 10-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 100 101 102 103 104

E[MeV]

10-5

10-4

10-3

10-2

10-1

100

101

102

103

104

D(E

)95

MeV

mb


neutrons


protons

pions Huhtinen

pions

electrons Summers

electrons






t =A

R=

A

σ middot L(613)












0

50

100

150

200

250

300

350

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

10

9N

EQ(

mm

2 middot1

07 s

)

p [GeVc]

Xe

Xe_max

N

N_max

p

p_max








































1 2 3 4 5 6 Total

015 GeVc+ 63 76 223 385 169 36 950

minus 45 90 210 384 171 34 932

05 GeVc+ 61 64 230 397 169 38 949

minus 54 71 224 409 166 36 945

10 GeVc+ 60 64 227 400 171 37 952

minus 56 69 224 408 166 38 950

15 GeVc+ 60 64 226 402 171 38 952

minus 57 68 224 408 166 39 952




663 Material Budget


XX0 =sumj

ρj middot LjX0j

(614)











〈XX0〉+ 2 middot 〈σθi〉 = 91 (615)




























664 Rate Estimation















Pixel part









Strip part
























671 Hit Resolution













Mean 109e-05

RMS 0002158

[cm]reco-xmcx-001 -0005 0 0005 0010

200400600800

100012001400


Mean 109e-05


Mean 1127e-05

RMS 0002886

[cm]reco-xmcx-001 -0005 0 0005 0010

200

400

600

800

1000


Mean 1127e-05

RMS 0002886 =1clust

Strip Resolution n


Mean -3275e-06

RMS 0001379

ndf 2

χ 4465 18



Sigma 00000074plusmn 00006904

[cm]reco-xmcx-001 -0005 0 0005 0010

500

1000

1500

2000

2500diffnxmul

Entries 51326

Mean -3275e-06

RMS 0001379

ndf 2

χ 4465 18





diffnymul

[cm]reco-xmcx-001 -0005 0 0005 0010

500

1000

1500

2000

2500

diffnymul gt1clust

Strip Resolution n

















681 Detector Setup





bull GEM 3 stations





bull Disc DIRC

bull Barrel DIRC

bull EMC crystals











Jψ rarr e+eminus

























m]microct [

0 500 1000 1500 2000

1

10

210

310







comb nonres





Vertex r and z cut 15004 689 1060 17255 194 543

4C kinematic fit 1058 25 0 1592 18 2




D+ Poca PRG KinVtx









]2M[GeVc17 18 19 2 21

210

310

Reco

Vtx-Cut

Vtx-Cut amp 4C-Fit


2 =87 MeVc4Cσ





D0 Poca PRG KinVtx



σz microm 884 949 902

D0

Poca PRG KinVtx



σz microm 884 941 893

Table 67 D0 and D0


6832 pprarr D0D0







m]microct [

0 200 400 600 800 1000 1200 1400

1

10

210

310




]2M[GeVc17 18 19 2 21

210

310

Reco

Vtx-Cut

Vtx amp 4C-Fit


2 =58 MeVc4Cσ




SB asymp 003 (617)







[GeVc] lab

p0 2 4 6 8 10 12 14 16 18

[m

b]

-210

-110

1

10

210

310

p totalp

6 prong

-3+3

0-3+3

-2+2K2





115





References

















































137





711 Pixel



















712 Strips




713 Integration





72 Safety












Acknowledgments





Task NamePanda MVD






CoolingDCS


Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q12009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019





143

Appendices

144

145

















bull FIFO buffering







146










147








1 α = arctan(

pitchthickness

)= arctan

(50microm300microm

)= 946

148










B1 B2 Device B3 B4


A 16 86 110 145 1855

B 90 100 110 120 130

C 65 85 110 139 159



149


A 56 53

B 16 16

C 34 34















150


GeVc mrad mrad

1 cm C p+ 295 102 101

2 cm C p+ 295 134 133


Air eminus 30 0423 0476

Air eminus 54 0243 0284







References



151


B1 Overview







bull FPGA Firmware


bull Readout PC




B2 Basic Concepts


Modularity






152























153







bull 16 LVDS inputs












bull LC-Display



B23 Firmware





154
















B3 Conclusion


References



155







156

157




x 883 694

y 694 883






158









159

D Alignment

D1 Introduction












160

















χ2 =sum

events

sumtracks

sumhits

∆2i σ

2i (D1)


161



xfit =

nsumi=1

aidi +

νsumj=1

αjδj (D2)



162









163







References








html







164

165

3D 3-Dimensional









BTeV B meson TeV









CNT Counter






CT Central Tracker

Cz Czochralski




DC Direct Current















FE Front-End










FZ Floating Zone




GBT GigaBit



166



GR Guard Ring










HV High Voltage





LD Laser Driver









MC Monte-Carlo



















PD Photo Diode




PMT Photomultiplier





RF Radio Frequency







167





SP Separation Power

SR Shift Register






TOF Time-of-Flight

ToPix Torino Pixel


TRx Transceiver


UI Unit Interval






VL Versatile Link





168

169

List of Figures












































170



322 ToPix v2 die 49










































364 SU8 strips 63














171















































172










































173











645 D0 and D0

decay lengths 133

646 D0 and D0



























174

175

List of Tables




































67 D0 and D0







Preface





113 Targets






124 Infrastructure


131 Basic Approach



21 General Overview



24 MVD Layout





32 Sensor





325 Production


331 Requirements


333 Front-End ASIC


335 ASIC Prototypes

34 Hybridisation

341 Bump Bonding


36 Module

37 Bus-Flex Hybrid



411 Barrel Sensors

412 Wedge Sensors


421 Requirements

422 Options


431 Architecture



44 Hybridisation

441 Overview




445 Test Assembly

5 Infrastructure





531 Introduction



54 Cables

541 Requirements

542 Signal Cable

543 Power Cable








563 Cooling Plant

57 DCS






632 Digitization

633 Noise Emulation




642 Track Finding

643 Track Fitting








663 Material Budget

664 Rate Estimation


671 Hit Resolution



681 Detector Setup






711 Pixel

712 Strips

713 Integration

72 Safety


Appendices










B1 Overview

B2 Basic Concepts



B23 Firmware


B3 Conclusion


D Alignment

D1 Introduction



List of Acronyms

List of Figures

List of Tables

PANDA Technical Design Report - MVD

Documents

Transcript of PANDA Technical Design Report - MVD