NUCLEARINSTRUMENTS

Nuclear Instruments and Methods in Physics Research A 338 (1994) 185-253

&METHODSNorth-Holland

IN PHYSICSRESEARCH

Section A

The DO detectorDO CollaborationS. Abachi h , M. Abolins q, B.S . Acharya ab , I . Adam g, S . Alm h , H. Aihara n,G. Alvarez 1 , G.A. Alves f , N. Amos P, W. Anderson m, Yu. Antipov W, S.H . Aronson °,R. Astur Z, R.E. Avery `, A . Baden °, J . Balderston k, B. Baldin W, J . Bantly d ,E. Barasch Z, J.F . Bartlett h , K . Bazizi e, T. Behnke Z, V. Bezzubov W, P.C . Bhat h,G. Blazey X, S. Blessing t , A . Boehnlein ad , F . Borcherding h , J . Borders X, N. Bozko W,

A. Brandt h, R. Brock q, A. Bross h , D. Buchholz `, V. Burtovoy W, J.M. Butler h ,O . Callot Z, D. Chakraborty Z, S. Chekulaev W, J . Chen b, L.-P . Chen n, W. Chen Z,

B.C . Choudhary e, J.H . Christenson h , D. Claes Z, A.R. Clark n, W.G. Cobau °,J . Cochran Z, W.E. Cooper h , C . Cretsinger X, D. Cullen-Vidal d , M. Cummings k ,D. Cutts d , 0.1 . Dahl n, B. Daniels °, K. De ac, M. Demarteau h , K. Denisenko h ,N . Denisenko h, D. Denisov W, S. Denisov W, W. Dharmaratna i, H.T . Diehl h ,M . Diesburg h , R. Dixon h, P . Draper ac, Y. Ducros Y, S. Durston-Johnson X,

D. Eartly h , P.H. Eberhard n, D. Edmunds q, A. Efimov W, J . Ellison e, V.D. Elvira h ,R . Engelmann Z, O. Eroshin W, V. Evdokimov W, S . Fahey q, G. Fanourakis ',M. Fatyga c, J . Featherly c, S. Feher Z, D. Fein b , T . Ferbel X, D. Finley h ,G. Finocchiaro Z, H.E. Fisk h , E. Flattum q, G.E. Forden b , M. Fortner s, P. Franzini g,S . Fuess h , E . Gallas ad , C.S . Gao h, T.L . Geld P, K. Genser h , C.E. Gerber h ,B . Gibbard e, V. Glebov aa, J.F. Glicenstein Y, B. Gobbi t, M. Goforth ', M.L . Good Z,

F. Goozen n, H. Gordon ', N. Graf c, P.D. Grannis Z, D.R. Green h , J . Green s,H. Greenlee h , N . Grossman q, P . Grudberg n, J.A. Guida Z, J.M. Guida `, W. Guryn c,N.J . Hadley °, H. Haggerty h , S . Hagopian ', V . Hagopian ', R.E. Hall e, S . Hansen h,J . Hauptman m, D. Hedin s, A.P . Heinson e, U. Heintz g, T. Heuring Z, R. Hirosky X,

K. Hodel X, J.S . Hoftun d , J.R . Hubbard Y, T. Huehn e, R. Huson ad , S . Igarashi h ,A.S . Ito h , E . James b , J . Jiang Z, K. Johns b , C.R . Johnson ae, M. Johnson h,A. Jonckheere h , M. Jones k , H. Jdstlein h, C.K. Jung Z, S. Kahn e, S . Kanekal g,A. Kernan e, L. Kerth n, A. Kirunin ', A. Klatchko ', B . Klima h , B . Klochkov W,

C. Klopfenstein Z, V. Klyukhin W, V. Kochetkov W, J .M . Kohli °, W. Kononenko v,J . Kotcher aa, I . Kotov `", J . Kourlas r, A. Kozelov `", E. Kozlovsky W, G. Krafczyk h ,K . Krempetz h , M.R . Krishnaswamy ab , P. Kroon c, S . Krzywdzinski h , S . Kunori °,S . Lami Z, G. Landsberg Z, R.E. Lanou d , P . Laurens q, J . Lee-Franzini Z, J . Li ac,

R . Li h , Q .Z. Li-Demarteau h , J.G.R. Lima f , S.L . Linn ', J . Linnemann q, R. Lipton h ,Y .-C. Liu t , D. Lloyd-Owen Z, F. Lobkowicz ', S .C . Loken n, S . Lokos Z, L. Lueking h,A.K.A . Maciel f , R.J . Madaras n, R. Madden ', E . Malamud h, Ph. Mangeot Y,I . Manning h , B. Mansoulié Y, V. Manzella Z, H.-S . Mao h , M. Marcin P, L. Markosky b,T . Marshall 1 , H.J . Martin 1 , M.I . Martin h , P.S . Martin h , M . Marx Z, B. May b ,A . Mayorov W, R. McCarthy Z, J . McKinley q, D. Mendoza a, X.-C . Meng h,

K.W. Merritt h , A. Milder b , A. Mincer r, N.K. Mondal ab , M. Montag c, P. Mooney q,M. Mudan r, G.T . Mulholland h, C. Murphy 1 , C.T . Murphy h , F . Nang d , M. Narain hV.S . Narasimham ab , H.A. Neal P, P. Nemethy r, D . Nesié d , K.K . Ng Z, D. Norman',L. Oesch P, V. Oguri f , E . Oltman n, N. Oshima h, D. Owen q, M. Pang m, A. Para h ,

0168-9002/94/$07.00 C 1994 - Elsevier Science B.V . All rights reservedSSDI0168-9002(93)E0560-F

186

DoCollaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

C.H . Park h , R. Partridge d , M. Paterno Z, A. Peryshkin h , M. Peters k , B . Pi q,H. Piekvrz i, Yu. Pischalnikov ', D. Pizzvto Z, A. Pluquet Y, V. Podstavkov W,

B.G . Pope q, H.B . Prosper h , S . Protopopescu c, Y.-K. Que h , P.Z. Quintas h ,G. Rahal-Callot Z, R. Raja h , S . Rajagopalan z, M.V.S . Rao a b , L. Rasmussen z,A.L . Read h , T . Regan a", S . Repvnd s, V. Riadovikov W, M. Rijssenbeek Z,

N.A. Roe ", P. Rubinov Z, J . Rutherfoord b , A. Santoro f , L. Sawyer ",R.D . Schamberger z, J . Sculli r, W. Selove °, M. Shev h , A. Shkurenkov ', M. Shupe b ,J.B . Singh ", V . Sirvtenko s, W. Smart h , A . Smith b, D. Smith e, R.P . Smith h ,G.R . Snow P, S . Snyder Z, M. Sosebee ae, M. Souza f, A.L . Spadafora ", S . Stvmpke q,R. Stephens ', M.L . Stevenson ", D . Stewart P, F. Stocker aa, D. Stoyanova w,H . Stredde h, K. Streets °, M. Strovink ", A . Svhanov w, A. Taketani h, M. Tartaglia h,

J.D . Taylor °, J . Teiger Y, G. Theodosiou ", J . Thompson z, S . Tisserant q,T.G . Trippe ", P.M. Tuts 9, R. Van Berg ', M. Vaz f , P.R . Vishwanath ab , A. Volkov W,

A. Vorobiev W, H.D . Wahl ', D.-C . Wang h , L.-Z. Wang h, H . Weerts q,W.A . Wenzel ', A. White a°, J.T . White ad, J . Wightman "d, S . Willis s,S .J . Wimpenny e, Z. Wolf ', J . Wvmersley aa, D.R . Wood h , Y . Xia q, D. Xiao ',P . Xie h , H . XU d , R. Yamada h , P . Yamin °, C . Yanagisawa z, J . Yang r, M.-J . Yang h ,C. Yoshikawa k , S . Youssef ', J . Yu Z, R. Zeller ae, S. Zhang P, Y.H. Zhou h , Q. Zhu r,Y.-S . Zhu h , D . Zieminskv 1 , A . Zieminski ', A. Zinchenko ', A. Zylberstejn Y

° Universidad de los Andes, Bogota, Colombiab University ofArizona, Tucson, AZ 85721, USABrookhaven National Laboratory, Upton, AT 11973, USA

d Brown University, Providence, RI 02912, USAe University of California, Riverside, CA, USALAFEX, Centro Brasileiro de Pesquisas Fisicas, Rio de Janeiro, Brazil

x Columbia University, New York, NY 10027, USA" Fermi National Accelerator Laboratory, Batavia, IL 60510, USA` University of Florida, Gainesville, FL 32611, USA' Florida State University, Tallahassee, FL 32306, USAk University of Hawaii, Honolulu, HI 96822, USA'Indiana University, Bloomington, IN 47405, USA`" Iowa State University, Ames, IA 50011, USALawrence Berkeley Laboratory, University of California, Berkeley, CA 94720, USA

° University of Maryland, College Park, MD 20742, USAP University of Michigan, Ann Arbor, MI 48109, USAQ Michigan State University, East Lansing, MI 48824, USANew York University, New York, AT 10003, USA

`Northern Illinois University, De Kalb, IL 60115, USA' Northwestern University, Evanston, IL 60208, USA" University of Panjab, Chandigarh, IndiaUniversity of Pennsylvania, Philadelphia, PA 19104, USA

`" Institute for High Energy Physics, Protuino, Moscow Region, Russian Federationx University of Rochester, Rochester, NY 14627, USAy. DAPNIA-CE Saclay, FranceState University of New York, Stony Brook, AT 11794, USA

°° Superconducting Supercollider Laboratory, Dallas, TX 75237, USAab Tata Institute of Fundamental Research, Bombay, India°` University of Texas at Arlington, Arlington, TX 76019, USAad Texas A&M University, College Station, TX 77843, USAae ZRL, Bristol, Rhode Island, USA

(Received 13 July 1993)

1. Introduction

Do Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

187

The DO detector is a large general purpose detector for the study of short-distance phenomena in high energy antiproton-pro-ton collisions, now in operation at the Fermilab Tevatron collider . The detector focusses upon the detection of electrons, muons,jets and missing transverse momentum . We describe the design and performance of the major elements of the detector, includingthe tracking chambers, transition radiation detector, liquid argon calorimetry and muon detection. The associated electronics,triggering systems and data acquisition systems are presented . The global mechanical, high voltage, and experiment monitoring andcontrol systems which support the detector are described . We also discuss the design and implementation of software and softwaresupport systems that are specific to DO .

The DO detector has been constructed to studyproton-antiproton collisions at v~'s =2 TeV in the Fer-milab Tevatron collider . The experiment was first pro-visionally approved in 1983 and the full conceptualdesign report [1] was prepared a year later. The primephysics focus of the DO experiment is the study of highmass states and large PT phenomena. These includethe search for the top quark, precision study of the Wand Z bosons to give sensitive tests of the standardelectroweak model, various studies of perturbativeQCD and the production of b-quark hadrons, andsearches for new phenomena beyond the standardmodel.

Fig. 1 . Isometric view of the DO detector.

This paper gives a description of the elements ofthe DO Detector and the systems for collecting datafrom it . Earlier reports [2-8] have summarized themain features and design goals of the experiment .

The DO detector was optimized with the followingthree general goals in mind :

- Excellent identification and measurement of elec-trons and muons.

- Good measurement of parton jets at large PTthrough highly segmented calorimetry with good en-ergy resolution .

- A well-controlled measure of missing transverseenergy (9T) as a means of signalling the presence ofneutrinos and other non-interacting particles.These principles derive from the observation that new

188

9

-s

Do Collaboration /Nuel. Instr. and Meth. in Phys. Res . A 338 (1994) 185-253

objects or phenomena typically have appreciablebranching ratios into states including leptons and jets,while the dominating QCD backgrounds have quitesmall leptonic branching fractions . The focus on partonjets is more important for deciphering the underlyingphysics processes than emphasis on the individual finalparticles emitted after hadronization.

The detector design which emerged had as its cen-tral design features :

- Stable, unit gain, hermetic, finely segmented, thickand radiation-hard calorimetry, based on detection ofionization in liquid argon . The inner radius of thecalorimetry was chosen to be small to accommodatethe desired depth and not compromise the surroundingmuon detector .

- Muon detection with thick magnetized iron ab-sorbers to provide sufficient momentum measurementand to minimize backgrounds from hadron punch-through.

- A compact non-magnetic tracking volume withinr= 75 cm with adequate spatial resolution and particu-lar emphasis on suppression of backgrounds to elec-trons .A cutaway isometric view of the detector is shown

in Fig. 1. The nested shells of the tracking and transi-

(m)

Fig . 2 . Elevation view of the DO detector. The scales are in meters .

tion radiation detectors, calorimetry and muon detec-tors are shown. The support platform on which thecore detector elements were assembled serves as thetransporter for the detector to and from the Do inter-section region . Much of the front-end electronics forthe detectors also rests on the platform . The elevationview of the detector shown in Fig. 2 shows the samedetector system shells and, in addition, the supportplatform in which detector electronics, cable connec-tions, and services for power, gas and cryogens arelocated. Fig. 2 also shows the Tevatron beam pipecentered within the detector and the lower energymain ring beam pipe used for preacceleration of pro-tons which passes above the detector center . The ca-bles from the detector elements pass over an articulat-ing bridge to the moving counting house (MCH), pro-viding sufficient separation of detector and MCH tokeep the latter in a radiation-safe environment outsidethe Tevatron shield wall. Digitized information pro-duced in the MCH is collected, zero-suppressed, andpassed over high-speed data highways to one of manydata acquisition processor nodes located in the controlroom area . Here, event building and event filteringfunctions are performed before passing an acceptableevent to the host computer for on-line monitoring and

recording on magnetic media . Events of sufficient in-terest are sent over the network to an "express line"disk storage for immediate reconstruction .

Subsequent sections of this paper discuss the goals,detector designs, electronics systems, and test perfor-mance data for the central detectors (Section 2),calorimeters (Section 3), and muon detection (Section4) . The various stages of triggering and data acquisitionare described in Section 5 . The main subsystems whichsupport the detector and electronics systems are sum-marized in Section 6 . The basic structure of the mainsoftware components of the experiment are outlined inSection 7 . We adopt a right-handed coordinate system,in which the z-axis is along the proton direction andthe y-axis is upward . The angles ¢ and 0 are respec-tively the azimuthal and polar angles (0 = 0 along theproton beam direction) . The r-coordinate denotes theperpendicular distance from the beam axes . Thepseudo-rapidity, 77 = -ln(tan(0/2)), approximates thetrue rapidity y = 1/21n((E +pz)/(E-pz)), for finiteangles in the limit that (m/E) - 0 .

2 . Central detectors

2.1. Central detector layout

The DO tracking and transition radiation detectorscomprise the central detector (CD) . The separate sys-tems are : (i) the vertex drift chamber (VTX), (ii) the

(D a

DO Collaboration INucl. Instr. and Meth. in Phys . Res. A 338 (1994) 185-253

transition radiation detector (TRD), (iii) the centraldrift chamber (CDC), and (iv) two forward drift cham-bers (FDC) . The VTX, TRD, and CDC cover the largeangle region and are arranged in three cylinders con-centric with the beams . The FDCs are oriented per-pendicular to the beams . Fig. 3 shows the layout ofthese detectors . The full set of CD detectors fits withinthe inner cylindrical aperture of the calorimeters in avolume bounded by r = 78 cm and z = ± 135 cm .

The design and optimization of the DO trackingdetectors was influenced by the absence of a centralmagnetic field . Without the need to measure momentaof charged particles, the prime considerations fortracking were good two-track resolving power, highefficiency, and good ionization energy measurement soas to distinguish single electrons from close-spacedconversion pairs . The TRD was included in order togain an additional factor of about 50 for rejection ofisolated pions beyond that given by the calorimeteralone . The scale for track spatial resolutions is set bythe need for primary z-vertex determination and bycalorimeter shower matching to be about 1 mm. Goodtrack-fitting efficiency and recognition of rr(K) - wdecay kinks benefit from as good resolution as can beattained . Some physics issues, such as the desire to tagseparated vertices arising from b-quark jets from topdecays, also stress the need for good position resolu-tion .

The transition between the large-angle trackingchambers with wires parallel to the beams and the

Central Drift

Vertex Drift

TransitionChamber Chamber Radiation

DetectorFig . 3 . Arrangement of the DO tracking and transition radiation detectors .

Forward DriftChamber

189

19 0

2.2. Vertex (VTX) chamber

Do Collaboration INucl. Instr. and Meth. in Phys . Res. A 338 (1994) 185-253

SENSITIVE REGION

WIRE BULKHEAD

small-angle tracking with wires perpendicular to thebeam was matched to the corresponding transitionbetween central and end calorimeters .

The CD detectors were designed to match the col-lider bunch-time interval of 3 .5 Rs . This time allowsrelatively long drift cells. Good two-track resolvingpower is obtained by employing a flash analog-to-dig-ital (FADC) conversion system for signal digitization inwhich the charge is sampled at - 10 ns intervals. Thisgives an effective detector granularity of 100-350 wm,but with relatively few channels (about 4200 wires and6080 total channels). In order to obtain robust mea-surement of the z-coordinate in the large-angle cham-bers, different methods are used for obtaining thatcoordinate in the VTX, TRD, and CDC. They includecharge division (VTX), helical cathode pads (TRD)and delay lines (CDC). The FDC also employs delaylines for measuring coordinates parallel to the wires.

More discussion of the design criteria for the CDdetectors is described in the DO Design Report [1] .Tests of prototype chambers [9-12] demonstrated thatthe design goals for DO could be achieved and aresummarized in section 2.7 .

The VTX chamber [12] is the innermost trackingdetector in DO. It has an inner radius of 3.7 cm (justoutside the beryllium beam pipe) and an outer activeradius of r = 16.2 cm . As shown in Fig. 4, there arethree mechanically independent concentric layers ofcells in the VTX chamber, each supported by thinG-10 bulkheads mounted on carbon fiber support tubes.

GAS BULKHEAD ; HV & SIGNAL FEEDTHROUGH -'

CABLE SUPPORT BULKHEADFig . 4. Plan view of the ends of the vertex chamber, showing the active chamber region, the support structure and the electronics

and feedthrough bulkhead .

Titanium tie rods connect each bulkhead to the nextinner carbon fiber tube to support the wire tensions . Ineach cell, eight sense wires provide measurement ofthe r-O coordinate . As indicated in Fig. 5, the inner-most layer (VTXO) has 16 cells in azimuth; the outertwo layers (VTX1 and VTX2) have 32 cells . Adjacentsense wires are staggered by ±100 Vm to resolveleft-right ambiguities ; the cells of the three layers are

Fig . 5 . r-0 view of a quadrant of the VTX chamber showingthe arrangement of the sense wires, grid wires and cathode

field shaping electrodes.

a

1000

1500

500

0

DO Collaboration /Nucl. Instr. and Meth. in Phys . Res. A 338 (1994) 185-253

191

-10 0 10Drift distance measured by PWC (mm)

Fig. 6 . Drift time vs drift distance in the VTX chamber forC02(95%)-ethane(5%) .

offset in 0 to further aid pattern recognition and tofacilitate calibration .

The sense wires are 25 wm NiCoTin [13] at 80 gtension. They have a resistivity of 1.8 kf1,/m and pro-vide a measurement of the z-coordinate from readoutsat both ends . The electrostatic properties of the cellare determined by the grounded planes of grid-wire oneither side of the sense wires planes and the outercathode field wires which shape the electric field in thecell . Coarse field shaping is provided by aluminumtraces on carbon fiber support tubes just beyond thefine field shaping wires. Field and grid wires are madeof 152 wm gold plated aluminum at a tension of 360 g.Further discussion of the electrostatics of the VTXchamber can be found elsewhere [11] .

To obtain good spatial resolution and track pairresolving power, the gas chosen for operation of theVTX is C0 2(95%)-ethane(5%) at 1 atm with a smalladmixture of H2O. Under our operating conditions,this gas is unsaturated ; good knowledge of the electricfield is necessary to obtain accurate velocities acrossthe cell . Fig. 6 shows the distance-time correlation forthe VTX chamber as measured in a test beam (thescatter of distance values is attributed to the impreci-sion of the beam particle position determination) . The

RADIATOR STACK

OUTER CHAMBER SHELL L 70 It GRID WIREFig. 7. The TRD detector showing the end of the polypropylene foil radiator section, the two 23 Wm windows separating radiatorand chamber, the conversion gap for TR X-rays, followed by the grid wire plane and the proportional wire section, and the outer

support structure.

19 2 Do Collaboration I Nucl Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

average drift velocity under normal DO operating con-ditions ((E) = 1 kV/cm) is about 7.3 wm/ns. Gas gainat the sense wires is approximately 4 x 10 4 . Studieshave shown that the addition of 0.5% H2O helpsstabilize the chamber operation in a high radiationenvironment [14]

2.3. Transition radiation detector (TRD)

The TRD occupies the space between the VTX andthe CDC and provides independent electron identifica-tion in addition to that given by the calorimeters .Transition radiation X-rays are produced when highly-relativistic particles (y > 103) traverse boundaries be-tween media with different dielectric constants [15] .The DO TRD consists of three separate units, eachcontaining a radiator and an X-ray detection chamber.Extensive studies and prototype tests were made tooptimize the TRD performance and to simplify theconstruction techniques [16-18].

The energy spectrum of the X-rays is determined bythe thickness of the radiator foils and the gaps betweenthe foils . The radiator section of each TRD unit con-sists of 393 foils of 18 wm thick polypropylene in avolume filled with nitrogen gas. The mean gap betweenfoils is 150 p,m; the fluctuations in the gap thicknessare comparable to the mean gap thickness. The gapsbetween foils are made by embossing a pattern in thepolypropylene sheets . This is achieved by laying thesheets over a polyethylene 2 x 2 cm mesh and heatingthe foil to 85'C for 5 min . The foil is cooled undersuction; the imprint of the mesh maintains the spacesbetween adjacent foils . The radiator foil is wounddirectly from the embossing facility onto its inner cylin-drical support. For the DO configuration, the transi-tion-radiation X-rays have an energy distribution whichpeaks at 8 keV and is mainly contained below 30 keV[18] .

The detection of X-rays is accomplished in a two-stage time-expansion radial-drift PWC mounted justafter the radiator . The X-rays convert mainly in thefirst stage of the chamber, and the resulting chargedrifts radially outward to the sense cells, where theavalanche occurs . Fig. 7 illustrates the construction ofthe TRD chamber with its conversion and amplifica-tion stages . The interaction length of the X-rays isenergy dependent, but the conversion usually occurs inthe first few millimeters of the conversion gap. Inaddition to the charge deposited by the transition-radi-ation X-rays, ionization is produced by all chargedparticles traversing the conversion and amplificationgaps . The X-ray conversions and ionization delta-raysproduce clusters of charge, which arrive at the sensewires over the full 0.6 l is drift-time interval . Fig. 8shows the arrival-time distribution of the charge gener-ated in the detection chamber due to transition-radia-

ôUG1Qs

w

0UOQr

w

60 F

50 Ir

40 'r

30 '~

20 '

20 40 60

Flash ADC bins (IOns)

Fig . 8. A typical time distribution of charge arriving at theTRD sense wire for traversal of a single 5 GeV electron .

tion X-rays and ionization from a single 5 GeV elec-tron . Both the magnitude and the time of arrival of theclusters of charge are useful in distinguishing electronsfrom hadrons. Fig. 9 shows the average ionizationobserved for electrons and for pions in the 5 GeV testbeam as a function of arrival time . The peak at shorttimes is due to ionization left in the amplificationstage; the excess energy for electrons at long times isdue to the conversion of transition-radiation X-raysnear the start of the conversion stage. The total energydeposited by 5 GeV electrons exceeds that depositedby pions of the same energy because of the transition-radiation X-rays and the relativistic rise of the specificionization .

The outer support cylinder of each of the TRDunits is a 1 .1 cm thick plastic honeycomb with fiber-glassskins . Kevlar end-rings support the cathode structures .The radiator section is enclosed in a carbon-fiber tubewith end flanges made of Rohacell with carbon-fiberskins. The radiator and detector volumes are separatedby a pair of 23 p m windows. Dry CO2 gas flows

8

6-

0 20 40 60 80 100 120

Flash ADC bins (IOns)

Fig . 9. Average charge collected on the TRD anodes as afunction of drift time for electrons and pions .

Do Collaboration INuel. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

through the gap between these two windows to keepthe nitrogen in the radiator from leaking into thedetector volume and polluting the recirculating cham-ber gas, a mixture of Xe(91%)/CH 4(7%)/CZH6(2%).These windows must be thin to minimize the attenua-tion of the low-energy transition-radiation X-rays . Theouter window is aluminized and serves as a high-volt-age cathode for the drift field in the conversion stageof the detection chamber. The windows are kept accu-rately cylindrical by maintaining small pressure differ-ences between the radiator, gap, and detector volumes.

The 15 mm conversion stage and the 8 mm amplifi-cation stage of the X-ray detector are separated by acathode grid of 70 Wm gold-plated tungsten wires. Theouter cathode of the amplification stage is made ofhelical copper strips deposited on Kapton foil. Theanodes of the amplification stage are 30 wm gold-platedtungsten wires (parallel to the chamber axis), separatedby 100 lLm gold-plated copper/beryllium potentialwires. Each TRD chamber has 256 anode readoutchannels . The outer chamber has 512 anode wires tokeep the basic electrostatic cell approximately"square", but the wires are ganged together two-by-twointo 256 readout channels . There are also 256 helicalcathode strips in each chamber, with pitch angles be-tween 24° and 47°. The thickness of the full TRD at9 = 90° is 8.1% of a radiation length and 3.6% of aninteraction length .

2.4. Central drift chamber (CDC)

The central drift chamber [9] provides coverage fortracks at large angles, after the TRD and just prior totheir entrance into the central calorimeter . The CDC isa cylindrical shell of length 184 cm and radii between49.5 and 74.5 cm . Fig. 10 shows an end view of theCDC. It consists of four concentric rings of 32 az-imuthal cells per ring . Each cell contains seven 30 wmgold-plated tungsten sense wires, read out at one end,and two delay lines located just before (after) the first(last) sense wires, each read out at both ends . Adjacentwires within the cell are staggered in 0 by ±200 ~Lm toremove the left-right ambiguity at the cell level. Inaddition, alternate cells in radius are offset by one halfcell to further aid in pattern recognition. The maxi-mum drift distance is - 7 cm.

The chamber construction [19] is modular; 32 sepa-rate identical modules are assembled into a singlecylinder as indicated in Fig. 10 . Each module is madefrom a set of Rohacell structural members ("shelves")covered with epoxy-coated Kevlar cloth and wrappedwith a double layer of 50 Wm Kapton. Grooves are cutinto each shelf at the azimuth corresponding to thesense wire locations to accommodate a Teflon tubecontaining a delay line . Resistive strips are screen-printed on the cathode surfaces to provide the field

193

Fig . 10 . End view of three segments of the central driftchamber.

shaping for the cells. The separate modules are posi-tioned within a single support cylinder . The innersurface is a composite carbon fiber/Rohacell tube tominimize conversions . The outer cylinder is 0.95 cmaluminum and serves as the main support for the fullCD . A precision-machined end plate contains ovalopenings into which the plugs containing the sense andpotential wires are fastened . In its final configuration,the CDCwire tension load is wholly transferred to thisend plate and the wire alignment is controlled by theend plate .

The delay lines embedded in the inner and outershelves of each cell propagate signals induced from thenearest neighboring anode wire ; measurement of thedifference of arrival times at the two ends permitslocation of the track along the z-coordinate . Surround-ing the first and last anode wires, an additionalgrounded potential wire is added to the usual pair ofpotential wires between anodes (see Fig. 10) to mini-

194

mize the signal induced upon the delay line from innersense wires. The inductive delay lines are constructedby winding a coil on a carbon fiber epoxy core . Propa-gation velocities are = 2.35 mm/ns; the ratio of delaytime to rise time is about 32 : l.

The CDC is operated with Ar(92.5%o)CH4(4%)-COZ(3%) gas with 0.5% HZO. Irradiation of a samplecell of the CDC using a 60Co source has shown that thechamber is stable for collected charges on the anodewires of up to 0.35 C/em . With a drift field of 620V/cm (in the region where dvdr;,/dE is negative) thedrift velocity is about 34 g,m/ns . The voltages on theouter sense wires are raised with respect to the innersense wires to induce larger delay line signals . The gasgains are 2 X 104 for the inner sense wires (6 X 104 forthe outer anodes) for the standard voltage settings of1.45 kV (1.58 kV). The grounded potential wires areprovided with current monitoring so that abnormalconditions can generate a voltage trip . Individual cellhigh voltages can be remotely turned off using a set ofrelays mounted in the detector platform .A single layer scintillating fiber detector for spatial

calibration was installed between the CDC and sur-rounding central calorimeter covering about 1/32 ofthe full azimuth. The 128 individual fibers are 1 mm indiameter and are aligned parallel to the beams; opticalreadout is made with a multi-anode photomultipliertube . Individual particle traversals are recorded withabout 30% efficiency ; the struck fiber can be used inconjunction with the measured CDC drift time to im-prove knowledge of the drift time vs distance relationand to regain the calibration constants rapidly uponchange of operating conditions .

2.5. Forward drift chambers (FDC)

Do Collaboration INucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

The forward drift chambers [9,20] extend the cover-age for charged particle tracking down to 0 = 5° withrespect to both emerging beams. As shown in Fig. 3,these chambers are located at either end of the con-centric barrels of the VTX, TRD, and CDC and justbefore the entrance wall of the end calorimeters . Theyextend to somewhat smaller radii than the large-anglechambers (r< 61 em) to allow passage of the cablesfrom the interior chambers.

Each FDC package consists of three separate cham-bers : the 0 module whose sense wires are radial andmeasure the (h coordinate, sandwiched between a pairof O modules whose sense wires measure (approxi-mately) the 6 coordinate. Fig. 11 indicates the sensewire orientations for each of these modules. The 45module is a single chamber containing 36 sectors overthe full azimuth, each with 16 anode wires along thez-coordinate traversed by particles. Each O moduleconsists of four mechanically separate quadrants, eachcontaining six rectangular cells at increasing radii. Each

081`8

Fig. 11 . View of the forward drift chamber O and 0 modules.

cell contains eight anode wires in z; the sense wires inthe three inner cells are at one edge of the cell so thatthe electrons drift in just one direction (removing theleft-right ambiguity) . Each O cell is equipped with onedelay line of identical construction to the CDC inorder to give local measurement of the orthogonalcoordinate . All adjacent anode wires (in z) of both 0and O modules are staggered by ±200 [Lm to helpresolve ambiguities . The upstream and downstream Omodules are rotated by 45° in 0 with respect to eachother. The maximum drift distance is 5.3 cm.

The 0 chamber electrostatic environment is deter-mined by a single grounded guard wire between an-odes . The cell walls are etched 25 wm aluminum stripson 125 wm G-10 to form the field shaping electrodes .The front and back surfaces are Kevlar-coated Nomexhoneycomb with copper traces on Kapton . The O mod-ule electrostatic cell employs two grounded guard wiresbetween adjacent anodes, similar to the CDC construc-tion . The front and back surfaces of the O cells areKevlar-coated Rohacell with copper traces on Kaptongiving the field shaping electrodes . The side walls are200 wm aluminum foil on Nomex honecomb .

The FDC chambers are operated with the same gasas the CDC, with similar values of the drift field andgas gain . The maximum drift time at the outer portionof the (P chamber is 1.5 ws .

2.6. Central detector electronics

The electronics for reading out signals from thecentral detectors are almost the same for all CD de-vices. There are three stages of signal processing : thepreamplifiers mounted directly on the chambers them-selves, the shapers located in the detector platform,and the flash ADC digitizers located in the MCH.

DO Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

Fig. 12 . Block diagram of the central detector signal digitization electronics .

Subsequent digital signal handling is considered part ofthe data acquisition system .

The sense wire, TRD cathode strip and CDC/FDCdelay line preamplifiers are based upon the FujitsuMB43458 quad common base amplifier [21] available ina surface mount package . The full CD requirement forthe DO tracking and TRD detectors is 6080 channels .

195

The preamplier gain is 0.3 mV/fC. Rise and fall timesare 5 and 34 ns respectively. Input noise is 2300 elec-trons for a detector input capacitance of 10 pF . Eighthybridized channels are mounted in a 28-pin DIP pack-age . Provision is made for test pulse charge injection tothe preamplifer inputs for calibration purposes .

The preamp output signals are carried via coaxial

Fig. 13 . FADC waveforms for test tracks in a CDC chamber cell . The top and bottom traces are for the delay lines (one end's signalis inverted in the display) .

CDCTILAYTFADC 4-NOU-1992 14 :27 Run 43097 Euent 947530-MAY-1992 01 :55

CDC Lager 3 Sector 4

~7V

A1143

207

L- -

131

2161

~l

-l38 -

ss

19 6 Do Collaboration /Nuel. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

cables of length about 15 m to the shaping circuits [22].The shaper contains a video amplifier, a two-zerothree-pole shaping circuit, and a cable driver . The gainand shaping components are mounted on separateheader boards so that the detailed shaping needs foreach detector and for cable matching can be accommo-dated.

The shaper output signals are transported over 45m on 75 SZ coaxial cables to the MCH where digitiza-tion occurs . Fig. 12 shows the block diagram for theCD digitization electronics, which consists of threemain sections . Design of this system was refined on thebasis of early studies using prototype tracking chambermodels and a flexible waveform digitization readoutsystem [10] . The input signals enter an analog bufferamplifer circuit [23] in which voltage offset and gaincorrections are made using stored digital informationfor each channel. The gain correction is such that largesignals are amplified by a factor of about 8.5 less thansmall signals, thus expanding the dynamic range of thedevice by a factor of about 3, or from 8 to an effective9.5 bits . Studies [10,24] on prototype chambers showedthat this extended response makes considerable im-provement in the dE/dx capability of the system .

The gain corrected signals are input to the flashADCsection, based on the SONY eight bit FADC partCX20116, operated at 106 MHz. Typical FADC out-puts are shown in Fig. 13. The digitized data from theFADC are stored in a FIFO circuit, pending a decisionto proceed with processing from the trigger system.

Since the full cell chamber drift times occupy up to256 FADC samplings and triggered events can occur atapproximately 200 Hz, the resulting data load would beof order 325 Mbyte/s if no data compression wereapplied . This load would exceed the capability of thedata acquisition data highways, and thus zero suppres-sion is required . It is performed in the final section ofthe digitization process, based on a custom-made appli-cation specific integrated circuit (ASIC) designed atFermilab [25] and manufactured by Intel, Inc. The zerosuppression (ZSP) circuit operates at 26 .5 MHz on 4byte words and hence is able to process data in realtime . It examines the sequence of digitized charges andadjacent FADC bucket charge differences . Digitizingsare saved between the leading edge and trailing edgeof a signal ; leading and trailing edges are defined byone of several algorithms based on digitized charges orcharge differences over threshold [26] . Thresholds aredetermined by real-time settable registers. Fig. 14 showsa schematic version of a pulse train from the FADCand illustrates the quantities (signal amplitudes Bi andadjacent bin differences D;) used for defining a se-quence of interesting FADC digitizations .

The three sections of the FADC/ZSP circuit arepackaged on a single VME 9U card . Each card con-tains 16 channels .

27-

2.7. Central detector performance

iJ s0 I I I I V 9 7 1 I I I

0 I i 1 18 I I

"

~- 9 .415 n s

~-r-LFL

lolsoltlllooollclo_

-l q .41sns

II~I ~LIC'Fig. 14. Sample waveform digitization from the FADC, showing the quantities used for determining pulse leading and

trailing edge .

All DO tracking detectors have been operated inboth particle beams and cosmic rays; descriptions andmore details on performance can be found elsewhere[9-12,20] . Correction of these detectors for gas contentand atmospheric conditions in the full experiment hasbeen assisted by the use of special monitoring "canary"chambers external to the detector. Separate canariesare operated on the full gas system for the VTX, TRD,and CDC/FDC [27] .

Fig. 15 shows the efficiency for finding hits in theVTX chamber as a function of high voltage. The morerapid rise to full efficiency is accomplished with lowerthresholds applied in the FADC hit-finding algorithms ;the penalty is an increase of the number of multiplehits found. Fig. 16 shows the hit-finding efficiency vshigh voltage for the FDC.

Spatial resolutions for the tracking chambers areshown in Figs . 17, 18 and 19, taken from test beamstudies with the final detectors . For the VTX chamber[12], resolutions were based on the deviation of aparticular hit from the expected value based on itsneighbors. Corrections for drift velocity variation acrossthe cell have been made in the VTX analysis . TheCDC and FDC resolutions are obtained from the devi-ation of a particular hit from the full track in thechamber. . In Figs . 18 and 19 for the CDC and FDC, thepoint for which resolution is measured was included in

UaUwwU40

bawI

0.9

0 .8

0 .7

0.60 .68 0.70 0 .72 0 .74

X, (esu/cm)I

I

I

]_ .I .2500 2550 2600 2650 2700

V, (volts)Fig. 15 . Hit finding efficiency as a function of high voltage forthe VTX chamber. Data points for both low and high thresh-

old settings are shown.

Do Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

0

0

I

I

I

. ® o ~ xo x x

xm

X

1200 1300 1400 1500 1600Sense Wire Voltage (V)

Fig . 16. Hit finding efficiency as a function of high voltage forthe FDC O and 0 chambers. The arrows indicate the operat-

ing points .

120

100

~+ 80

0b

60

40

20

197

L

I

I

I

I

I

I0 2 4 6 8 10 12

Drift distance (mm)Fig. 17 . Position resolution vs position for the sense wires of

the VTXchamber.

the fit ; if excluded, the resolutions degrade by about10%. Resolutions of (typically) 50 wm are seen for theVTX chamber; the resolutions in the CDC and FDCvary between 150 and 200 Wm.

Pulse pair resolutions are important in the non-magnetic environment of DO. Figs . 20 and 21 show theefficiency for resolving two hits as a function of theirtime separation . These data result from off-line super-position of single track test beam data . For the VTX,90% efficiency for two-hit resolution is achieved for0.63 mm separation ; for the FDC (or CDC), 90%efficiency is reached for a =2 mm separations [28] .

The delay lines used in both CDC and FDC havebeen studied using cosmic ray data. The coordinatealong the delay line is determined from the time differ-ences at the two ends. Propagation velocities weredetermined in a special calibration fixture and cor-rected using test beam data . Single delay line positionresolution for the FDC is measured to be o-Z = 4 mm.The CDC delay lines have been shown [19] to haveo, = 2 mm when fitting to the eight delay lines along atrack. The z-resolution from charge division in theVTX was measured in a prototype chamber [111 to bebetter than 1% of the wire length or = 1 cm .

It is important for the DO tracking chambers todistinguish an unresolved overlay of two tracks (e .g .from y-conversions) from a single track. Fig. 22 shows

1 .0

0.8Ua 0 .6U

w 0 .4

0 .2 Theta Chamber

0 .0 I I 11 .0 . .. . . . .. . .. . .. .. . . . . . . . . . . . .. . . . . . . . . .. . .

0 .8Uv 0.6

w 0.4.

w0.2 Phi Chamber

0 .0 I I ~ I

198

the ionization distributions for one and two tracktraversals of a single cell (two track events are made byoffline overlay of single track events). In Fig. 22, theionization signal is taken to be the sum of the smallest70% of the hits participating in the track; this proce-dure reduces the effects of delta-ray production [27] .Studies using the FDC [20] show an energy loss resolu-tion of 13.3% for single tracks . Studies using the CDCand VTX show rejection factors for two overlappedtracks in the range 75-100 for 98% efficiency forretaining single tracks .

The ability of the TRD to discriminate electronsfrom hadrons has been studied in some detail in testbeams [16,18] and the results compared to a detailedMonte Carlo calculation [29] . Measures of the differ-ence between electron and hadron include the totalcollected charge on anode wires, the number of charge

i=~t 360

°m 270

2253N 180

45

x360c

cn 270a225

3180

0L0

O90

45

0L0

DO Collaboration INucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

Test beam data

f

r

"

+ 4

_

-------

MonteMonte Carlo

_. ._ __ .______.___._ . .__ .__ . __ ._________ .._ . ._

Test beam data

clusters, and the position of clusters determinedthrough the time of arrival at the anodes . Fig. 23 showsthe observed energy distribution, summed over all threeTRD layers, for pions and electrons . Use of totalenergy alone allows a rejection factor of about 10 forpions while retaining 90% of the electrons. The distri-bution of the likelihood function,

3In (PQ(E,)1P,(EJ)

=1is shown in Fig. 24 ; the probablities for a given energyin layer i are based upon the observed total energy andnumber and position of clusters . Fig. 25 shows the pionrejection factor as a function of electron inefficiencyfor particle incidence angles of 90° (normal) and 50°.Pion rejections of approximately 50 are found for 90%efficiency for electrons in this analysis using the likeli-

4

4

5

Drift Distance (cm)

5

Drift Distance (cm)

Fig. 18 . Position resolution vs position for the inner and outer sense wires of the CDCchamber.

3 . Calorimeters

DO Collaboration INucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

Fig . 19. Position resolution vs position for the sense wires ofthe FDC O and 4) chambers .

hood function . The rejection factors are better at non-normal incidence, in part because of the increasedpath length and in part due to space charge distortionsat normal incidence induced by pile-up of all ionizationalong a particular track .

3 .1 . Calorimeter design issues

The calorimeter design [1,30] is crucial for the opti-mization of the DO detector . Since there is no centralmagnetic field, the calorimetry must provide the energymeasurement for electrons, photons and jets. In addi-tion, the calorimeters play important roles in the iden-tification of electrons, photons, jets and muons, and inestablishing the transverse energy balance in an event .Early in the design phase, the decision was taken toemploy liquid argon as the active medium to samplethe ionization produced in electromagnetic or hadronicshowers . This choice was supported by the unit gain ofliquid argon, the relative simplicity of calibration, theflexibility offered in segmenting the calorimeter intotransverse and longitudinal cells, the good radiationhardness, and the relatively low unit cost for readoutelectronics. Factors weighing against the choice of liq-uid argon included the complication of cryogenic sys-tems ; the need for relatively massive containment ves-sels (cryostats) which give regions of uninstrumentedmaterial ; and the inaccessibility of the calorimetermodules during operation.

TUcU

w

S

0 50 100 150 200 250 300Separation (ns)

0.0 0 .5 1 .0 1 .5 2.0Separation (mm)

Fig . 20. Efficiency for resolving two hits in the VTX chamberas a function of their separation distance .

Since provision must be made for some degree ofaccess to the central detectors within the calorimetercavity, more than one vessel is necessary . We chose thesolution shown schematically in Fig . 26 in which acentral calorimeter (CC) covers roughly 10771 _< 1 anda pair of end calorimeters, (ECN (north) and ECS(south), extends the coverage out to I rl I = 4 . Theboundary between CC and EC was chosen to be ap-proximately perpendicular to the beam direction ; this

0 .8

0 .6

0 .4

0 .2

0 .0

199

0

2

4

6Separation (mm)

Fig . 21 . Efficiency for resolving two hits in the FDC chamberas a function of their separation distance .

Theta Chamber

300 1 .0

2000.8

O 100O

N NU

0w

0 .6Phi Chamber c .

M300 a

Ia~m 0.40 200 .

a

ô 100 0.2

00 1 2 3 4 5

Drift Distance (cm) 0.0

200

450

ô

a

,400z

350

300

250

200 F

150 E

100

50

Do Collaboration /Nucl. Instr. - and Meth . in Phys. Res. A 338 (1994) 185-253

1 MIP CUT

0

0.4

0.8

1 .2

1 .6

2

2.4

2.8

3.2 - - 3 .6 -

- 4Minimum ionizing units (MIPS)

Fig. 22 . 70% truncated mean of dE/dx for one and two tracks seen in the full CDC chamber.

choice was shown to introduce less degradation inmissing transverse energy (,~,) than one in which theECs nest within the CC shell with a boundary approxi-mately parallel to the beams.

The dimensions of the calorimeters were set by theconstraints imposed by the size of the experimental

Fig . 23. Distribution of total energy observed on the threeTRD layers for both electron and pion traversals.

2 MIP

hall, the need for adequate depth to ensure goodcontainment of shower energy, the requirements ofmagnetic measurement of muon momenta outside thecalorimeter, and the need for sufficient tracking cover-age in front of the calorimetry. The resulting designhas three distinct types of modules in both CC and EC :

=I -

j-7.5

-5

' -2.5

0.

2.5

5.

7.5

10 .Likelihood

Fig. 24 . Distribution of the likelihood value calculated forelectrons and pion traversals of the three TRD layers.

DO Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45Eelectron

Fig . 25 . Rejection factor for 5 GeV pions vs electron inefficiency for track angles of 90° (normal incidence) and 50° .

These data are obtained using the likelihood parameter .

an electromagnetic section (EM) with relatively thinuranium absorber plates, a fine-hadronic section withthicker uranium plates and a coarse-hadronic sectionwith thick copper or stainless steel plates . Inclusion ofthe coarse sections allows sampling of the end ofhadronic showers while keeping the density high (andhence outer radius small) . Except at the smallest an-gles in the EC, several (16 or 32) modules of each typeare arranged in a ring . The modular design providesunits of workable size without creating undue compli-

cations from degraded response near module bound-aries . At i7 = 0, the CC has a total of 7.2 nuclearabsorption lengths (AA); at the smallest angle of theEC, the total is 10.3AA .A typical calorimeter unit cell is shown in Fig . 27 .

The electric field is established by grounding the metalabsorber plate and connecting the resistive surfaces ofthe signal boards to a positive high voltage (typically2.0-2 .5 kV) . The electron drift time across the 2.3 mmgap is = 450 ns . The gap thickness was chosen to belarge enough to observe minimum ionizing particlesignals and to avoid fabrication difficulties .

Different absorber plate materials were used indifferent locations . The EM modules for both CC andEC used nearly pure depleted uranium [31] ; the thick-nesses were 3 and 4 mm respectively (production diffi-culties with the larger EC plates necessitated thethicker plates) . The fine hadronic module sections have6 mm thick uranium-niobium (2%) alloy [31] . Thecoarse hadronic module sections contain relatively thick(46 .5 mm) plates of either copper (CC) or stainlesssteel (EC) . Electrical connections to all absorber plateswere made by percussive welding of thick niobiumwires to the edges of the plates . This technique provedto be simple and very reliable . The density of these(ground) connections is about 1 for every 2-4 signalconnections .

To avoid breakdown across the gap, absorber elec-trodes must have relatively smooth surfaces . The sur-

Fig . 26. Isometric view showing the central and two end calorimeters .

- CENTRALCALORIMETER

ElectromagneticFine Hadronic

Coarse Hadronic

201

202 Do Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

face treatment of the uranium plates evolved as manu-facturing experience was gained . Although oxidation inair was not generally a problem, care had to be takento avoid oxidation of the uranium through the reactionU + 2H20 _' U02 + 2H 2 which is strongly suppressedby the presence of oxygen [32] . For the earlier produc-tion of the CC uranium, surfaces were cleaned bymechanical abrasion . Subsequently, EC plates werecleaned using jets of high pressure hot water. Mostsurfaces remained free of excessive oxidation, althoughsome unexplained centimeter-scale patches of oxideappeared on a small number of plates . These wereremoved when possible since the oxide is sufficientlyconducting to give unwanted current draw in a module .

Signal boards for all but the EM and small-anglehadronic modules in the EC were constructed by lami-nating two separate 0.5 mm thick G-10 [33] sheets .Each signal board had one surface coated with highresistivity (typically 40 MSZ/11) carbon-loaded epoxy[34] . One of the inner surfaces was left with bare G-10on the uncoated side ; the other sheet, originally cop-per-clad, was milled into the pattern desired for thesegmented readout. Several such pads at approxi-mately the same vl and (b are ganged together in depthto form a readout cell . Details of signal ganging tomake the readout cells vary from module to module .

The two smallest angle modules in each EC havethe added problem that even small gaps betweenneighboring azimuthal sectors would give undesireddead regions. Thus, the signal boards for these mono-lithic modules were made from multilayer printed cir-cuit boards . The outer surfaces were coated with the

Absorber Platesr L Ar Gaps

1 Unit Cell

same resistive epoxy as for the other signal boards .Etched pads on an interior surface give the desiredsegmentation . Signal traces on another interior surfacebring the signals to the outer periphery. The pad andtrace layers are connected by plated-through holes.The signals from these multilayer boards in the electro-magnetic (ECEM) and the small angle hadronic (ECIH)modules were ganged together along the depth of themodules using solder-tail header connectors and Kap-ton printed circuit lines. In the ECIH module, theganged signals were brought to the rear plate of themodule using twist and flat ribbon cable.

In all other modules, ganging connections weremade using insulation displacement connectors (IDCs)and solid wires. This technique proved to be veryreliable : there are no observed IDC failures with ap-proximately 500000 connections in the installed mod-ules .

The pattern and sizes of readout cells were deter-mined from several considerations. The transverse sizesof the cells were chosen to be comparable to thetransverse sizes of showers: - 1-2 cm for EM showersand - 10 cm for hadronic showers. The variables moredirectly useful for physics are Ovl and 0¢; the scale isset by the typical size of parton jets, OR =1/077 2 + A02 - 0.5 . Segmentation finer than this is use-ful in probing the shape of jets . Longitudinal subdivi-sion within the EM, fine hadronic and coarse hadronicsections is useful since the longitudinal shower profileshelp distinguish electrons and hadrons. Variations incell capacitance cause changes in the pulse rise time,but the sampling of signals occur at fixed time intervals

Cu Pads

Fig. 27. Schematic view of the liquid argon gap and signal board unit cell .

DO Collaboration INucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

synchronized to the accelerator bunch crossing time(At= 3.5 ws) . This implies that each readout cell, withassociated cabling, should have a capacitance that doesnot exceed 5 nF .

The final design was chosen to have a "pseudo-pro-jective" set of readout towers, with each tower subdi-vided in depth. The term pseudo-projective refers tothe fact that the centers of cells of increasing showerdepth lie on rays projecting from the center of theinteraction region, but the cell boundaries are alignedperpendicular to the absorber plates . Fig. 28 shows aportion of the DO calorimeter segmentation pattern.There are four separate depth layers for the EMmodules in CC and EC . The first two layers are typi-cally 2 radiation lengths (X,) thick and are included tohelp measure the longitudinal shower developmentnear the beginning of showers where photons and -rr°sdiffer statistically . The third layer spans the region ofmaximum EM shower energy deposits and the fourthcompletes the EM coverage of approximately 20X,.The fine hadronic modules are typically segmentedinto three or four layers ; coarse hadronic modules areganged into one or three layers . Typical transversesizes of towers in both EM and hadronic modules areAil = 0.1 and t10 = 2 ,rr/64 = 0.1 . The third section ofEM modules is twice as finely segmented in both 77 and

¢ to allow more precise location of EM shower cen-troids . Inter-module gangings are provided in the earlystages of the front-end electronics (prior to digitiza-tion) in order to join segments of cells which cross ECand CC module boundaries .

3.2. Central calorimeter

The central calorimeter (CC) shown in Fig. 29 com-prises three concentric cylindrical shells . There are 32EM modules in the inner ring, 16 fine hadronic (FH)modules in the surrounding ring, and 16 coarsehadronic (CH) modules in the outer ring. The end viewof the CC in Fig. 29 shows that the EM, FH and CHmodule boundaries are rotated so that no projectiveray encounters more than one intermodule gap.

Fig. 30 shows a schematic view of a CC module .Each is made by loosely stacking the alternating ab-sorber plates and signal boards within a stainless steelwelded box structure consisting of perforated side skins,somewhat thicker front and back walls, and thick ma-chined end plates . Delrin spacers inserted throughholes in the signal boards extend 2.3 mm on one sideand 1.5 mm on the other side of the boards . Adjacentspacers have opposite long and short segments so thatthe boards have some freedom to flex in response to

20 3

Fig. 28. Schematic view of a portion of the DO calorimeters showing the transverse and longitudinal segmentation pattern . Theshading pattern indicates the distinct cells for signal readout. The rays indicate the pseudorapidity intervals seen from the center of

the detector .

204

mechanical and thermal stresses . The absorber platesare prevented from movement along the long axis ofthe modules by keys on the side skins.

The end plates serve as the support members forthe modules in the ring structure . Large fasteners arebolted on, linking adjacent endplates into the cylindri-cal rings. Support for the CC is transmitted to theunderlying detector platform through four feet whichpass through the cryostat vessel to a supporting cradle .Flexible support stanchions transfer the load from theinner cold vessel to the outer room temperature vac-uum vessel [35] .

DO Collaboration /Nucl. Instr. and Meth . in Phys. Res . A 338 (1994) 185-253

Fig . 29. (a) End view of the central calorimeter . (b) r-z viewof half of the central calorimeter .

Signal and GroundConnctions

3.3. End calorimeters

Fig . 30. Schematic view of the CC module construction .

HV

The CCEM modules have four longitudinal gang-ings of signals of approximately 2.0, 2 .0, 6 .8, and 9.8X0 ,with transverse segmentation as described above . Theganged signals from the pads making up a particularreadout cell are collected on special readout boardslocated just behind the signal boards of a given depthand are transmitted along the length of the module toconnectors at the end plates . The total number ofsignals for the 32 modules is about 10400, spanning24t1R = 0.1 towers along the 260 cm length. A fullmodule comprises 20.5X0 and 0.76AA and weighs 0.6metric tons .

The CCFH modules have three longitudinal gang-ings of approximately 1 .3, 1 .0 and 0.9AA . The CCCHmodules contain just one depth segment of 3.2 kA . Thesignal collection scheme is similar to that for the CCEMmodules. The CCFH (CCCH) modules weigh 8.3 (7 .2)metric tons; the full rings provide about 3500 (770)signals . The total weight of the CC modules and theirsupport structure is 305 metric tons, with an additional26 metric tons contributed by liquid argon.

The two mirror-image end calorimeters (ECN andECS) contain four module types as shown in Figs. 26and 28. To avoid the dead spaces in a multi-moduledesign, there is just one EM module and one innerhadronic (IH) module, shown schematically in Fig. 31 .Outside the EM and III, there are concentric rings of16 middle and outer (MH and OH) modules . As seen

DO Collaboration /Nucl. Instr. and Meth. in Phys. Res . A 338 (1994) 185-253

in Fig . 32, the azimuthal boundaries of the MH andOH modules are offset to prevent cracks through whichparticles could penetrate the calorimeter.

The ECEM modules [36,37] contain four readoutsections (0.3, 2 .6, 7 .9, and 9.3Xo) with outer radiivarying between 84 and 104 cm and inner radius of 5.7cm . The material of the cryostat wall brings the totalabsorber for the first section up to about 2Xo . Thealternating absorber plates and signal boards of theECEM modules were stacked on both sides of a thickstainless steel support member whose thickness in radi-ation lengths is equivalent to a uranium plate . Eachcircular absorber plane consisted of three separateuranium plates. As indicated in Fig . 33, a collection oftie-rods and spacers position the assembly of absorberplates and signal boards from the surface of a "centralsupport plate" ; the weight of the absorber plates issupported from the central support plate by a centralsupport tube. Each ECEM module weighs 5 metrictons and provides 7488 signals .

The two ECIH modules are cylindrical, with innerand outer radii of 3.92 and 86.4 cm . The fine hadronicportion consists of four readout sections, each contain-ing sixteen 6 mm semicircular uranium plates ('.'AAeach). Alternate plates have their boundary rotated by90° to avoid through-going cracks. The coarse hadronic

portion has a single readout section containing thirteen46 .5 mm stainless steel plates (4.1AA ) . The full ECIHmodule weighs 28.4 metric tons and provides 5216signals .

Each of the ECMH modules has four fine-hadronic(uranium) sections of about 0.9AA each and a singlecoarse-hadronic (stainless steel) section of 4AAA . Sig-nal boards are similar to those used in the CC . Signalsare brought to the azimuthal boundaries where theyare ganged by short jumper cables to make thepseudo-projective cells . Each ganged signal is trans-ported to a connector located at the rear of the mod-ule . ECMH module weights are 4 .3 metric tons ; thereare 1856 signals in an MH ring .

The ECOH modules employ stainless steel platesinclined at an angle of about 60° with respect to thebeam axis, as shown in Fig . 28 . Module weights are 5 .5metric tons; a full ring has 960 signals .

The ECOH module ring forms a support platformfor the remaining EC modules. Inconel 718 and 304stainless steel studs are inserted into ears on the stain-less steel plates near the front and rear of OH modulesto fasten together adjoining modules. The ring ofECMH modules resting within the OH ring supportsboth the ECIH and ECEM module . The OH ring loadis transferred via a beam and strap assembly to four

Fig . 31 . View of ECEM. The readout boards form complete disks with no azimuthal cracks. The ECIH construction is similar.

205

206 DO Collaboration INucl. Instr. and Meth . in Phys . Res . A 338 (1994) 185-253

bosses and through the cryostat on flexible stanchionsas for the CC calorimeter described above. The overallEC calorimeter weighs about 238 metric tons .

3.4 . Intercryostat detectors and massless gaps

As shown in Fig. 28, the region 0.8< 1771 < 1 .4contains a large amount of uninstrumented material inthe form of cryostat walls, stiffening rings, and moduleendplates. The material profile along a particle pathvaries rapidly with rapidity through this region . Tocorrect for energy deposited in the uninstrumentedwalls, we have built two scintillation counter arrayscalled intercryostat detectors (ICD) that are mountedon the front surface of the ECs (see Fig. 28). Each ICDconsists of 384 scintillator tiles of size An = AO = 0.1exactly matching the liquid argon calorimeter cells . Inaddition, separate single-cell structures called masslessgaps also shown in Fig. 28 are installed in both CC andEC calorimeters . One ring with standard segmentationwas mounted on the end plates of the CCFH modules

(320 channels); additional rings were mounted on thefront plates of both ECMH and ECOH modules (192channels). Together, the ICD and massless gaps pro-vide a good approximation to the standard DO sam-pling of EM showers.

The ICD tiles are constructed from Bicron BC-414scintillator [38] into which three or four 3 mm groovesare milled . Two millimeter diameter bundles of 200Wm polystyrene wavelength-shifting scintillating fibersare epoxied into these grooves to collect the light andcarry it through a 180° bend to phototubes mountedjust behind the tile . Fig. 34 shows the arrangement offibers in a scintillator tile . Tests have shown that thecollected light is uniform across the face of a tile towithin about 10%. The tiles are packaged into light-tight boxes containing three tiles at fixed .0 as shown inFig. 34b.

The ICD readout uses 1.3 cm diameter phototubes[39] . All tubes were subjected to extensive testing andburn-in to reduce subsequent failures and to providematched triplets to be installed in a given box with a

Fig . 32. End view of the end calorimeter showing the ECEM at small radii and the surrounding rings of ECMH and ECOHmodules.

3.5 . Cryogenic system

DO Collaboration /Nucl. Instr. andMeth . in Phys. Res. A 338 (1994) 185-253

common high voltage supply . The overall combinationof scintillator, wavelength shifter and phototube gaveapproximately 20 photoelectrons per minimum ionizingparticle for the crucial larger ?7 boxes . Since the typicalnumber of particles crossing a given tile in a typicalhighPT two-jet event is of order 25, the signals fromthe ICD are quite large . Relative calibration of theICD response is maintained using optical fibers whichdistribute light from a UV laser to each separate tile .

Three separate double-walled stainless steel cryo-stats enclose and support the DO calorimeters, asshown in Fig. 28 . The outer diameter of these vessels is518 cm ; CC and EC cryostats are approximately 306and 263 cm long, respectively . Forty layers of superin-sulation fill the vacuum space between inner and outershells . The torospherical ends of each cryostat can becut off (and subsequently rewelded) to give access tothe interior volume for module assembly. Each cryostathas four 20-cm diameter ports for the passage of signal

cables . In addition, each vessel has two smaller portsfor the penetration of the high voltage, temperatureand purity monitoring cables . Four feet support eachcryostat and its modules upon an external cradle . Thesefeet [35] are specially constructed from low thermalconductivity laminations and are also capable of flexingto accommodate the displacements induced duringcalorimeter cool-down .

The main ring beam passes through the threecryostats near the outer radius of the vessels; theTevatron beams pass through the EC vessels on thecenterline . Bellows are incorporated in these penetra-tions to accommodate the thermal and differentialpressure motion of the ends of the vessels to whichthey are welded .

The front heads of the EC cryostats are speciallydesigned to provide both strength and minimum mate-rial in front of the EM module . A small radius convex(thin) spherical section is joined to a large radius(thick) concave annulus at a stiffening ring locatedbetween ECEM and ECMH modules . The EC cryostatshave inner cylindrical surfaces attached to the rest of

207

Fig . 33 . Layout of the ECEM signal pads for the four layers in depth . The upper portion of the ECEM indicates the uranium platedivisions and the location of the support bolts .

208

__

iiiii

7 .850iii

iii

J---- II-_____-5.816--____J

DO Collaboration /Nucl. Instr. and Meth . in Phys. Res . A 338 (1994) 185-253

the vessel through a flexible stainless steel bellows; thispermits the needed flexibility of motion needed nearthe beam pipe . Special rotating cryogenic supply tubeswere designed to allow retraction of the ECs to giveaccess to the tracking detectors.

Each calorimeter has three finned-tubing coolingcoils which are located near the top of its cryostat .Independent liquid-nitrogen flow control is providedfor the topmost coil, intended for steady-state opera-tion ; the two symmetrically located lower coils are usedonly for calorimeter cool-down. Temperature monitorsare mounted at many locations on modules and cryo-stat walls . A maximum 50 K difference between anytwo monitors is imposed during cool-down . Cooling toliquid argon temperature (78 K) occurs in approxi-mately 10 days . Once cool, argon is introduced in partby means of liquid transfer from a 20000 gallon argonstorage tank at approximately the same elevation asthe calorimeters and in part by condensation of gaseousargon using the cooling coils . Rapid transfer of the fullload of liquid between storage dewar and cryostats ispossible (in about 8 h) using modest gas pressures .Such empty and fill operations occur after each moveof the detector between collision hall and assemblyhall . The system has been designed to warm up inabout six weeks .

100 micron Fibersin 2 nn bundles

__

i 0 .394(a)

--------- 6.568-_______J

Sc int i llator :1 cm thick

An inventory of 20 000 gallons of liquid argon ismaintained in the cryostats and in the storage dewar .Argon purchases specify an Oz contamination of lessthan 0.6 ppm . Each cryostat and storage dewar isequipped with four argon-purity monitoring detectors[40] . Each such device contains a radioactive ß-source,triggered so as to give full traversal of a standard argongap by the ß-particle . In addition, there is a cellequipped with an a-source . The dependence of o.- andß-signals on applied voltage is a sensitive indicator ofthe equivalent oxygen content of the argon [41] . Inde-pendent measures of oxygen and nitrogen contamina-tions are obtained using standard gas monitors . Wehave observed that the response of test cells, as well asof the calorimeters, drops less than 0.5% over thecourse of many months of operation .

3 .6. Calorimeter readout electronics

Signals from the calorimeter modules are broughtto the four cryostat feedthrough ports by speciallyfabricated Tefzel [42] insulated 30 fl coax cables . Thesecables are connected to multilayer printed-circuitfeedthrough boards located above the liquid level ineach feedthrough port . Eight T-shaped, 27-layerfeedthrough boards penetrate each port and provide a

Fig. 34 . (a) The ICD the design showing the placement of the readout optical fibers . (b) Schematic representation of the ICD boxeswith three tiles, fiber, and phototube combinations .

DO Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

re-ordering of signals betw6en the module-orientedinput side and the v1-0 order appropriate to analysison the output side .

Short cables connect the output from the feed-through boards to charge-sensitive hybrid preamplifiers[43-46] mounted in four enclosures on the surface ofeach cryostat, near the ports . A schematic diagram ofthe calorimeter preamplifiers is shown in Fig. 35 . Asingle 2SK147 Toshiba j-FET with gm - 0.05 H-1 isused at the input . The equivalent noise density ise. = 0.46 nV/ Hz . Preamplifiers with two differentgains were made (equivalent full scale gains of about100 and 200 GeV) to provide full dynamic range re-sponse . Variation in the gains of the preamps wascontrolled to within about 0.5% .

Output signals from the preamplifiers are trans-ported on twisted pair cable some 30 m to baselinesubtractor (BLS) shaping and sampling hybrid circuits[43-46]; input signals are integrated (430 ns) and dif-ferentiated (33 Ws) . Fig . 36 shows a schematic diagramof the BLS . Signals from portions of logical cells thatstraddle the CC-EC boundary are merged at the BLSinput. At the input to the BLS, a portion of the signalis extracted with = 100 ns rise time and added intotrigger towers of î1v1 = A0 = 0.2 for use in event selec-tion . The main signals are sampled just before abeam-crossing and 2.2 Vs after; the difference is pro-vided as a do voltage proportional to the collectedcharge . Two storage capacitors for each channel allowdouble buffering at the analog level . Fast baselinerestoration occurs in a few ws so as to minimize theeffects of event pile-up . Fig . 37 shows the equivalentsignal after BLS sampling and difference.

15K

Fig . 35 . Schematic of the calorimeter preamplifier.

Depending on signal size, the BLS outputs can beamplified by 1 or by 8 so as to reduce the dynamicrange requirements of the subsequent digitization . Abit is set to record the chosen amplification; there isspecial provision for forcing a specific amplification (orboth) for calibration purposes. The BLS outputs aremultiplexed 16-fold onto the crate backplane, and eachchannel is sent in serial time slices over 50 m twistedpair cables from the detector platform to the MCH.

The 24-channel 12-bit ADC circuits [43-46] in theMCH, together with the x 1 or x 8 amplifier allows anoverall dynamic range of 2 15 . Each time-slice of eachchannel is digitized in about 10 ws, yielding a totaldigitization time of 160 Ws for 384 signals. Gain param-eters are set so that about 3.75 MeV of depositedenergy correponds to one least count. Minimum ioniz-ing particles deposit between 8 (EMl) and 90 (FH1)MeV in the layers of the calorimeters . Channels whichhave an absolute difference of signal and pedestalbelow an adjustable threshhold can be suppressed fromthe readout buffer.

Both single-channel random noise (electronics anduranium radioactivity) and multi-channel coherentnoise have been measured in beam tests and in the DOhall . The single-channel noise can be characterized as2000 + 3100 x C (nF) electrons. For many channels,the total random noise varies with ~N, where N is thenumber of channels included ; the total coherent noisevaries typically as N. A useful measure of coherent'noise is the number of channels that can be summedbefore the coherent noise exceeds the random noise .In the DO environment this number is typically > 5000channels [47] .

51

51-6V

20 9

15K0.22uF

YI 1 U

15K 15K

'

0 .22uFWK.4. .7uF

210

Calibration of the various components of the elec-tronics is accomplished by a precision pulser [48] whichinjects charge through a large accurately known resis-tor at the front end of each preamplifier . The pulsedistribution system has been carefully designed to de-liver equal pulses at each input. A range of attenuatorsettings can be programmed to allow calibration overthe full dynamic range of the system . Precision andstability of the calibration system have been demon-strated to be better than 0.25% .

3.7. Calorimeter performance

The Do calorimeters have been tested in a varietyof ways . Prototype studies in test beams [49-52] haveverified performance goals and led to the optimizationof the design . Subsequent beam tests involved combi-

Do Collaboration INucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

nations of the CCEM/CCFH and ECMH modules[53-56] as well as the ECEM/ECIH modules[36,37,57-61] . The full CC calorimeter was tested usingcosmic rays before the detector moved to the beam line[62] . Details of these measurements can be found inthe references .

The response of the ECIH module as a function ofhigh voltage for 100 GeV pious and electrons is shownin Fig . 38 . For comparison, the response of the test cellto ß-particles is also shown . The three distributions aresimilar and the slope of the plateau is small, character-istic of oxygen contamination below 1.0 ppm . Smalldifferences between electrons and the ß-cell are seendue to effects of induced voltages on the moduleuranium surfaces . Electron and pion responses candiffer due to the presence of heaviliy ionizing compo-nents in hadron showers .

OUrPUr

OUTPUT

Fig . 36. Schematic diagram of the BLS signal shaping and trigger pickoff circuit (above) and the BLS sample and hold circuit(below) .

3.0

2.5

2.0

1 .5

1 .0

0 .5

0.0

-0 .5

Fig. 37. Equivalent signal shape after BLS filter, doublesample and difference .

Extensive studies of the performance of moduleswere made using pions and electrons with energiesbetween 10 and 150 GeV [63]. Fig. 39 shows theresponse to 100 GeV pions in the ECEM/ECIH. Thelow energy excess seen in the distribution is due toupstream interactions in the beam and cryostat materi-als and can be eliminated by requiring that < 0.4 GeVbe observed in the first EM layer . Electron distribu-

EÔNNCOn

NNE0.9

V

DO Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

Time (us)

0.8

0.7

0.6400 800 1200 1600

ca)w

20 40 60 80 100 120 140Reconstructed Energy (GeV)

Fig. 39 . Response to 100 GeV pions, with EM1 <0.38 GeV(shaded), and with no cut on EMI (unshaded).

tions are essentially Gaussian . The response to bothelectrons and pions is linear with beam energy towithin 0.5% . The longitudinal profile of the energydeposited by pions is shown in Fig. 40 .

2000 2400 2800 3200Calorimeter Gap Voltage (kV)

Fig. 38 . Signal response for electrons, pions and 13 test cell as a function of high voltage .

212

t

C

c17dC

NQ

21a)CW

0 05 -

_ 0 03

(o,,IE) 2 = C 2 +S 2IE +N2/E2,

Do Collaboration /Nuel. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

13 DataGEANT/GHEISHA Monte Carlo

+ Parametrization

Depth in Interaction lengthsFig . 40. Longitudinal energy depositions for pions between 50and 150 GeV. Test beam data is shown with the boxes,

together with Monte Carlo simulations .

Energy resolution for electrons and pions obtainedafter subtraction of pedestals, and after corrections forgain variations and beam particle momentum (of theorder of 1%), are shown in Figs. 41 and 42 . Therelative resolution as a function of energy for theECEM and ECMH modules can be parametrized as

0

50

100

t50

Beam Momentum (GeV/c)

Fig. 41 . Fitted values of Q/E vs energy for electrons in theECEM module .

0 .20

U1

w 0 .10N

O

0 .05

or I

I

I

I

I

I

I

I

I

I

I

I

0

50

100

Beam momentum (GeV/c)

Fig. 42 . Fitted values of o-/E vs energy for electrons andpions in the ECMH module .

where the constants C, S, and N represent the calibra-tion errors, sampling fluctuations and noise contribu-tions respectively . We obtain C = 0.003 ± 0.002 andS= 0.157 ± 0.005

GeV for electrons in ECEM andC = 0.032 ± 0.004 and S = 0.41 ± 0.04

GeV for pionsin ECMH. The noise contribution is dominantly due touranium radioactivity, and is consistent with our previ-ous discussion of random and coherent noise.

The calorimeter position resolution is important foridentification of backgrounds to electrons from thenear-overlap of photons and charged particles. Fig. 43shows the position resolution for 100 GeV electronsdetermined from comparison of the test beam trackand the energy-weighted position of the shower deter-mined solely from ECEM layer 3 . This resolution variesbetween 0.8 and 1.2 mm over the full range of impactpositions; the position resolution varies approximatelyas E1/ 2 .

0

- Average resolution

Resolution at pad edge

50

L00Heam Momentum IGoV

Fig . 43 . Position resolution for electrons in the third layer ofECEM . The pad size is 2.5 cm .

z 160ww

140

120

100

80

60

40

20

0

Do Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

MUON PULSE HEIGHT DISTRIBUTION (WITH BACKGROUND)

FINE HADRONIC LAYER 1

0 50 100 150 200 250 300 350 400

ENERGY DEPOSITED (ADC COUNTS)

Fig . 44 . Energy deposition by cosmic ray muons in the CCFHLayer 1 . Noise contributions, shown with dark shading, arededuced from signals observed far from the cosmic track. The

data are plotted with a 3-or suppression .

A good measure of the sensitivity of the calorime-ters is their response to muons. Fig. 44 shows theenergy seen in the CCFH layer 1 in a tower centeredcosmic ray track. The noise contribution indicated isdeduced from the signal observed in an equivalentroad perpendicular to the track. A three standard

1 .20

1 .15

1 .05

1 .000 50 100 150

Beam Momentum (GeV/c)Fig . 45 . Ratio of electron to pion response in the ECEM+ECIH modules after correction for transverse energy leakage .

Test beam data and Monte Carlo simulations are shown.

d

C7

zw

60

50

40

x 30wzw 20

10

60

50

40

30

20

10

0

213

-2

0

2X(CM)

Fig. 46 . Scatter plot of energy vs position across the intermod-ule boundary between two CCEM modules for 25 GeV elec-trons. (a) Before correction for energy loss in unintrumented

material . (b) After correction .

deviation suppression on the noise signal is applied tothe data.

The resolution and linearity obtainable in thecalorimeter is closely related to the ratio of response ofelectrons and pions [64] . Fig. 45 shows this ratio for theECEM/ECIH data . Corrections have been been ap-plied for effects due to out-of-time event pileup, earlyshowering, and energy deposition outside the 077 X AOregion used to sum the energies . The e/ , rr responseratio falls from about 1.11 at 10 GeV to about 1.04 at150 GeV.

Correction procedures have been developed to ac-count for energy lost within uninstrumented materialin module walls and cryostats . Fig. 46(a) shows a scat-ter plot of total energy (EM + hadronic) vs positionacross the boundary between two adjacent CCEMmodules near ?7 =0. The dip at x = 0 corresponds tothe intermodule crack. Using the full energy correla-tion matrix [65] and the observed energy deposits inthe neighboring modules, a predicted correction [66]for the energy deposits in the dead material can bemade . Fig. 46(b) shows the distribution after applyingthis correction . A similar technique has been devel-oped [67,68] for using ICD, massless gaps, and regularcalorimeter deposits to correct for the losses in thethick CC and EC cryostat walls in the range 0.8< 1771< 1.4 . Energy depositions with and without theICD/massless gap corrections are shown in Fig. 47 .

214

4. Muon detection

4.1 . Muon system design

The DO muon detection system consists of fiveseparate solid-iron toroidal magnets, together with setsof proportional drift tube chambers (PDTs) to measuretrack coordinates down to approximately 3° . The pur-pose of this system is the identification of muonsproduced in p-p collisions and determination of theirtrajectories and momenta. Since muons are measuredafter most of the debris of electromagnetic andhadronic particle showers is absorbed in the calorime-ters, muons can be identified in the middle of hadronjets with much greater purity than electrons .

9 400a)w

w

DO Collaboration INucl. Instr. and Meth. in Phys . Res. A 338 (1994) 185-253

300

200

100

800

600

400

200

0

50

Since the bend in the toroids (B - Bo¢) is approxi-mately in the r-z plane and the Tevatron interactionregion is long (o,, = 30 cm), it is necessary to measuremuon directions both before and after the bend . Aclosely spaced set of measurements of the muon trackis provided before the toroid and yields the entrancepoint; two sets of measurements separated by 1-3 mafter the toroid yield the exit direction. The incidenttrajectory is determined from a combination of theprimary interaction point, the track seen in the centraltracking and the first muon chamber track vector .Muons can also be seen in the calorimeter . A compari-son of incident and exit muon directions provides thebend angle in the toroid, and hence the momentum .Multiple Coulomb scattering in the iron toroids limits

- CC+EC+ICD+MG. . CC+EC only

x-100

150

200

(a)

Energy (GeV)

- CC+EC+ICD+MG. . CC+EC only

100

150

200

(b)

Energy (GeV)

Fig. 47 . Distribution of observed calorimeter energy for single 100 GeV pions, before and after applying corrections based on ICDand massless gaps at (a) 77 =1 .15 and (b) v7 =1 .25 .

the relative momentum resolution to >_ 18% up to thelimit imposed by the bend coordinate resolution in thePDTs . With final precise alignment of the muon cham-bers, three standard deviation determination of thesign is expected for PT _< 200 GeV/c at 7 = 0 andPT < 30 GeV/c at 1771 = 3.3 .

The central toroid (CF) covers the region

71 _< 1

and two end toroids (EFs) cover 1 < 1711 < 2.5 . Thesmall-angle muon system (SAMUS) differs somewhatfrom the larger angle system ; the WAMUS toroids fit inthe central hole of the EF toroids and cover 2 .5 < 1 ?7 1

<_ 3 .6 . Fig . 48 shows the elevation view of the DOdetector with the five toroids and their associated PDTlayers indicated . The muon system is quite thick, asshown in Fig . 49 and affords a clean environment foridentification and momentum measurement of highPTmuons [69] . Fig . 50 shows the geometrical acceptanceof the muon chamber system . Some loss of acceptanceoccurs near 1771 = 0.75 due to the CF/EF transition .Apart from gaps caused by various support elementsand the need to provide access corridors to the detec-tors, the coverage for muons is hermetic. The mini-mum momentum required for a muon to emerge fromthe iron toroids varies from about 3.5 GeV/c at 71 = 0

to about 5 GeV/c at larger 71 .

Do Collaboration INucl. Instr. and Meth. in Phys . Res . A 338 (1994) 185-253

The wide angle muon system (WAMUS) providesmeasurements for all muons traversing the CF andmost of those which cross the EF toroids. The WA-MUS system consists of 164 distinct proportional drifttube chambers with sizes up to 2.5 x 5.8 m2 . TheWAMUS chamber wires are oriented along the pri-mary B-field direction to give accurate measurement ofthe bend coordinate . Each end of the WAMUS systemcontains three stations with three planes per station .Fig . 51 shows the disposition of these chambers .

4.2 . Muon toroids

The CF toroid [70,71], shown in perspective in Fig .52, is a square annulus 109 cm thick and weighs 1973metric tons ; it is centered on the Tevatron beam lines .The inner and outer surfaces of CF are 317 .5 and 427cm from the beams respectively. In order to give accessto the inner portions of the detector, the CF is built inthree separate pieces. A middle (bottom) section is a150 cm wide beam fixed to the detector platform whichalso provides the support for the enclosed calorimetersand tracking detectors . The CF is completed by twoC-shaped shells which can be moved perpendicular tothe beams to allow access to interior detectors . The

215

Fig . 48 . Elevation view of the DO detector showing the five muon toroids and the approximate dispositions of the A, B, and Clayers of proportional drift tubes.

216

wzad

Û0a

w0

ae

0 .

0

02

Do Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

Fig . 49 . Variation of the thickness of the DO calorimeters and muon toroids as a function of polar angle .

0 .5

1

1 5

2

2.5

3

3 .5

IqIFig . 50. Geometric acceptance of the muon system vs 77 for allmuons passing through two and three layers of muon PDTs .

INTERACTION LENGTH (M

2

2 .5

3

4

mating surfaces at top and bottom include 4.8 and 3mm stainless steel spacers respectively, which reduceremanent field and facilitate separation of the sections .Twenty coils of 10 turns each carry currents of 2500 Aand excite internal fields of 1.9 T. Since the coils arenot symmetrically placed on the CF (there are no coilson the central beam), the fringe fields are moderatelystrong and exceed 0.01 T near the central beam . Leak-age represents 3.7% of the total CF flux.

The two EF toroids are located at 447 <

z

<600cm. Their outer surfaces are at a perpendicular dis-tance of 417 cm . The EFs have a 183 cm square innerhole centered on the Tevatron beams. The main ringbeam passes through a 25 cm diameter hole in the EFs.Eight coils of eight turns each carry 2500 A (in serieswith the CF) and excite fields of up to about 2 T. EachEF toroid weighs 800 metric tons .

Within the inner hole of each EF toroid there is aseparate SAMUS toroid . The SAMUS toroids, weigh-ing 32 metric tons each, have outer surfaces at 170 cmfrom the beam axis and 102 cm square inner hole . Two

Do Collaboration /Nucl. Instr. and Meth. in Phys . Res . A 338 (1994) 185-253

coils of 25 turns each carry currents of 1000 A. Fig. 53shows a magnetic field contour plot for EF andSAMUS; the field vectors in the two magnets arealigned in the same directions. The EF/SAMUS com-binations can be retracted along the beam line byapproximately 1 m in order to provide access to thedetectors around the interaction point.

and the support beams in the detector platform.Fig. 52 . Perspective view of the two halves of the CF toroid

Fig. 51 . Exploded view of the muon chamber layout .

217

A pair of special tungsten-lead collimators wereconstructed to fill the space between the SAMUStoroid and the Tevatron beam pipe . These collimators

Fig . 53 . Contours of magnetic induction for the EF/SAMUScombination for the case that the B in the two magnets are

the same .

21 8

Fig . 54 . Extruded aluminum section from which the "B" and"C" layer PDT chambers are constructed. The "A" layerchamber extrusions are similar, but have four cells instead of

three.

intercept beam halo and collision products at smallangles and thereby reduce the rate of hits in the smallangle nation system . Multiplicities of clusters observedin the A, B and C stations of SAMUS are typicallyabout 9, 5, and 11 per event.

4.3. Wide angle muon chambers

The WAMUS chambers [72] are deployed in threelayers : the "A" layer before the iron toroids and the"B" and "C" layers after the magnets. The distancebetween B and C layers is >_ 1 m, so that a goodmeasurement of the direction of the muon after themagnet can be made . There are four PDT planes in anA layer chamber and three PDT planes in the B and Clayer chambers. The cell structure for the PDTs is thesame for all WAMUS chambers ; the 164 individualWAMUS chambers differ mainly in the number ofcells in depth (three or four) and width (between 14and 24), and in their length (between 191 and 579 cm).

The WAMUS PDTs are formed from aluminumextrusion unit cells as shown in Fig. 54 . The extrusionsare cut to the appropriate lengths and press fit into anidentical neighbor . Fig. 54 shows a relative transverseoffset between the planes of chambers that allowsbreaking of the left-right drift-time ambiguity. Cath-ode-pad strips are inserted into the top and bottom of

DO Collaboration /Nucl. Instr. and Meth . in Phys . Res. A 338 (1994) 185-253

KWA

Fig. 55 . A single cell of the PDT chambers showing theequipotential surfaces surrounding cathode pads and anode

wire .

each unit cell and the anode wire is held near thecenter of the cell, as shown in Fig. 55 . The maximumdrift distance is 5 cm. The 50 [Lm gold-plated tungstenanode wires are held at 300 g tension by a special plugmounted in an aluminum cap extrusion covering theends of the chamber. The holes containing the wiresare drilled with a precision of ± 130 ~tm. Wire sag over610 cm is 0.6 mm. The aluminum extrusion itself isoperated at ground potential with the cathode padsand anode wires held at +2.3 kV and +4.56 kVrespectively. The total number of WAMUS PDT cellsis 11386.

The coordinate (~) along the wire direction is mea-sured by a combination of cathode pad signals andtiming information from the anode wires. Wires fromadjacent cells are jumpered at one end, and signals aretaken for the pair from the other ends in order tosimplify access to chamber electronics . A coarse indica-tion of the 6 coordinate can be obtained by measuringthe time difference for a particular anode signal fromthe two ends of the paired wire . Two hits per wire-pairare accommodated to allow for S rays . The At mea-surement determines 6 with a precision varying from10 to 20 cm along the wire .

Fine resolution in ~ is obtained using informationfrom the cathode pad signals . The upper and lower

Fig. 56 . Layout of the two separate electrodes comprising the PDT cathodes. The signals on inner and outer segments of thediamond pattern are independently measured .

cathode planes shown in Fig . 55 are made from twoindependent electrodes, as shown in Fig . 56, formingthe inner and outer portions of a repeating diamondpattern whose repeat distance is 61 cm . The two innerpads of a given cell are added and read independentlyof the sum of the outer pads . Subsequent calculation ofthe ratio of sum and difference of inner and outersignals gives a measure of the 6 coordinate, modulothe approximately 30 cm half-wavelength of the pat-tern . The pointer for the correct cathode pad solutionis given by the At measurement . The diamond patternsfor adjacent planes of PDT are shifted in 6 by about1/6 of the repeat distance to minimize the ambiguitiesencountered near the diamond pattern extrema . Theresolution achieved in a given chamber is approxi-mately ±3 mm.

Cathode pads were fabricated from copper-cladGlasteel (polyester and epoxy copolymer sheets withchopped glass fibers) [73]. The copper cladding wasrouted into the desired diamond pattern by a computercontrolled milling machine . After cutting to the appro-priate lengths, the pad surfaces facing the active por-tion of the cell were covered with 50 wm thick Kaptontape to ensure electrical isolation from the chamberextrusion.

The chambers are operated using the trinary gasmixture Ar(90%)/CF4(5%)/COz(5%) for which thedrift velocity is about of order 6.5 cm/ws, but variesacross the cell with changing E [74,75] . Tests of thisgas have shown a nearly linear time-distance relation.Typical leak rates of completed chambers were about0.005 standard cubic feet per hour.

4.4. Small angle muon chambers

DO Collaboration INucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

The layout of the SAMUS system was shown in Fig.51 . The A station precedes the SAMUS toroid; the Band C stations are between the toroid and the begin-ning of the low-beta quadrupole for the DO insertion .The SAMUS stations cover 312 x 312 CMZ square ar-eas perpendicular to the exit beams . An interior 61(86) cm square hole permits passage of the beam pipein the A and B (or C) stations . Each station consists ofthree doublets of 29 mm internal diameter cylindricalproportional drift tube chambers [76] . These doubletsare oriented in x, y and u directions (u is at 45° withrespect to x, y) . Individual PDTs in a doublet form aclose packed array with adjacent tubes offset by one-half a tube diameter . There is a total of 5308 tubes inthe SAMUS system .

The SAMUS PDTs are constructed from 3 cmexternal diameter stainless steel tubes with individualend plugs for provision of gas and electrical connec-tions, as shown in Fig . 57 . The anode wire is 50 Wmgold plated tungsten and is tensioned with 208 g . Thesagitta for these wires is 0.24 mm over a length of 3 .1

Fig . 57. SAMUS drift tubes : (1) HV supply wire ; (2) insulatingtube ; (3) 30 mm stainless steel outer tube ; (4) silicon rubber;(5) epoxy ; (6) groove ; (7) anode wire ; (8) copper pin; (9) end

cap ; (10), (11) solder; and (12) ground wire .

m . The system is operated with CF4(90%)/CH4(10%)gas whose average drift velocity is 9 .7 cm/ws . Theresulting maximum drift time is 150 ns . The time ofarrival is approximately linear with drift distance. Sur-vey of the tubes within a station was made to a preci-sion of 0.3 mm.

The SAMUS A and B stations are mounted fromthe EF toroids on the same support structures as theWAMUS EF A and B chambers . The C station issupported on the free-standing truss which holds thesmall angle portion of the WAMUS C layer chambers .

4.5. Muon readout electronics

219

Because of the large area over which the muonchambers are dispersed, much of the signal processingelectronics for the PDTs is located on the chambermodules . Signal shaping, time-to-voltage conversions,latching of struck cells, chamber monitoring and multi-plexing of signals for efficient transport to the digitiz-ers are all performed locally . Digitizers and triggerelectronics reside in the MCH.

Each of the WAMUS chambers has its set of elec-tronics boards housed in a specially constructed enclo-sure on the side of the chamber body . These includeseparate "motherboards" (fast signal shaping) for eachsix cells of each plane of the chamber, one "corner-board" per chamber (multiplexing and signal-drivers tothe digitizers and triggers), and one "monitor" boardper chamber (hardware-status monitoring) .

220 Do Collaboration INucl. Instr. and Meth . in Phys. Res . A 338 (1994) 185-253

The motherboard contains a set of hybrid circuitswhich provide signal shaping and time-information en-coding . For each cell, the sum of upper and lowercathode-pad signals from the inner and outer portionof the diamond pattern is brought to a hybrid circuitcharge-sensitive preamplifier (CSP). This hybrid is verysimilar to that used for the calorimeter preamplifier.Each CSP is followed by a baseline subtractor circuit(BLS), again like those for the calorimeter ; the BLSperforms before and after sampling of the signal andstores the difference on one of two output capacitors .Signals from the anode wires arrive at the ends of apair of wires jumpered at their far ends ; these areamplified and discriminated in a pair of hybrid circuits(2WAD) . The 2WAD outputs are supplied to twotime-to-voltage hybrids and two time-difference ("De-Ita-time") to voltage hybrids which provide analog in-formation on time and time difference for up to twohits on the wire pair . The final element on the mother-board is the pad latch which records the PDT cells thathave had their cathode pads struck, based on theinformation present in the pad BLS hybrids.

Each chamber contains a cornerboard which col-lects the signals from the motherboard. The pad latchinformation is multiplexed by a factor of 4 and sentacross the long cables from the chambers to the MCH.In addition, the latch bits are ORed on the corner-board to provide summary information on muon activ-ity and majority logic indicating 1, 2, 3, or 4 hits in thechamber. The analog information from the pad BLSsand the time and Delta-time circuits is multiplexed andtransmitted in 96 time slices to digitizers in the MCH.Provision is made in the cornerboards for future inclu-sion of signals from an array of scintillation counters tobe mounted on the outer portion of the detector. Thecornerboard also contains the test pulser for front-endelectronics calibration .

The monitor board associated with each chamber isconnected to the general experiment monitoring ser-vices by the token ring described in Section 6.7 . Itprovides the local interface to the operating character-istics of the chamber, including temperature, gas flow,voltages, and currents . The monitor board is also usedto set thresholds for latching pad cells and for testpulser amplitudes .

For each SAMUS PDT anode signal, the front-endelectronics provide a card with the amplifier, discrimi-nator, time-to-voltage converter, and a latch indicatinga struck tube . Under the supervision of a SAMUScontrol board, these signals are multiplexed and sent tothe MCH in a similar fashion to the WAMUS informa-tion . The monitor board for SAMUS is identical tothat for WAMUS. The latch bit information is thebasis for formation of small angle muon triggers .

Digitization of the muon chamber cathode pad sig-nals and voltage-encoded time information is per-

formed in the MCH using 12-bit ADC circuits whichare variants of those used for the calorimeter digitiza-tion . There are 50 920 analog elements in the muonsystem .

4.6. Muon chamber performance

Tests of the muon chambers have proceededthrough several phases. Early studies [69] measured thepenetration probabilities for particles through magne-tized steel within roads around incident muons andhadrons and helped establish the thicknesses of thetoroids. Subsequent studies were conducted using WA-MUS PDT chambers in test beams [77] and cosmic rays[78,79] . Early tests employed gas mixtures of Ar(90%)/COZ(10%) with drift. velocity 4.6 cm/Ws. Subsequentoperation with the faster gas Ar(90%)/ COZ(5%)/CF4(5%) whose drift velocity is about 6.5 cm/ws hasshown qualitatively similar performance characteristics .

Response of the PDT pad latches to cosmic rays asa function of anode potential is shown in Fig. 58 ; anoperating range of 100-200 V is observed for each ofseveral potentials applied to the cathode pads. Thecorresponding pad signal amplitudes are shown in Fig.59 .

In a test stand employing a set of proportional wirechambers with resolution +0.3 mm to define cosmicray tracks in a WAMUS PDT, the track coordinatecould be determined independently from the PDT timemeasurement. Fig. 60 shows scatter plots of drift time

Muon Drift Chamber

Anode Voltage

kV (Wire)

Fig . 58 . Rate of cosmic ray tracks seen as a function of thePDT anode potential for a set of different cathode pad

potentials .

tnCO

U0

900

800

700

-600

C

a)500

~, 400

300

200

100

a

DO Collaboration INucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

4 .2

4.4

4.6

4 .8

Wire Voltage (kV)Fig . 59. Pulse height observed on PDT cathode pads as afunction of anode potential for three different values of

cathode potential.

and position for tracks incident at 0° (normal) and 45°.Although the drift time relation is approximately lin-ear, detailed fits showed non-linearities of about ± 1mm due to the variations in electric field within the

900

800

700

600

500

400

300

200

100

-5

-2.5

0 '

2.5

5

-5

-2.5

0 -

2.5

5

Drift distance (cm)

Drift distance (cm)Fig. 60 . Scatter plot of muon PDT drift times vs position asdetermined by a separate proportional wire chamber system,

for angles of incidence of 0° and 45° .

400

~ 350

>> 300u- 250Ov 200

É 150

Z 100

50

0

Drift Time Spatial Resolution

PWC-PDT Full Cell (mm)

PWC-PDT Short Segment (mm)

221

Fig. 61 . Deviation of a muon PDT hit coordinate from thepredicted position from the PWC telescope, for full lengthchambers and short test modules. The measurements aremade using a PWC system which adds 0.29 mm in quadrature

to the deviations which are not subtracted in these data .

cell . Fig. 61 shows the comparison of PDT coordinate(determined from the linear approximation) and thepredicted location based on the PWC information .After unfolding the PWC resolution, this linear ap-proximation gave the PDT drift-coordinate resolutionof about ±0.53 mm. Employing the fitted non-linearfunction to the distance-time curve of Fig. 60 reducedthe PDT resolution to ±0.3 mm, which is close to thediffusion limit. Fig. 62 shows the variation in PDTresolution with angle of incidence. Studies of thechamber efficiency as a function of position within theunit cell showed nearly full efficiency except in the 2mm window at the cell extrusion wall .

Cosmic ray studies were also conducted to measurethe 6 coordinate resolution along the PDT wires. Be-cause of attenuation and dispersion effects, the Atmeasurement taken from the time of arrival of pulsesat the ends of adjacent wires has a resolution which isdependent on the 6 coordinate . The At measurementerror was found to be ±9 cm near the ends of thewires furthest from the amplifier/ discriminators and± 23 cm nearest the electronics . This is sufficient toresolve the ambiguity of the repeating diamond pat-tern . The ratio of difference and sum of outer andinner cathode pads is shown in Fig. 63 . Fig. 64 showsthe distribution of deviations in the 6 coordinate asdetermined from the cathode pad ratios using the

222

2 .4

Eb

0 .4

na0Ow

c7

00

Do Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

Incident Angle Relative to Normal (degrees)

Fig . 62 . Variation of muon PDT resolution as a function ofthe track angle of incidence.

POSITION ALONG WIRE (rrm)

Fig. 63 . Dependence of the ratio (Qa - Qb)l(Qa + Qb ) withthe coordinate 6 along aPDT wire . The quantities Qa and Qb

refer to signal charges on outer and inner portions of thediamond shaped cathodes .

e0

20

12c 60NW

E:3 40Z

30

20

10

0 r - r-,--1-7-71-10 -8

Muon Drift Chamber Pad Response

Pad Resolution (mm)

o = 1 .6 mm

Fig. 64 . Resolution in coordinate ~ along the wire from thecathode pad ratio signal.

PWC. The fitted resolution for the ~ coordinate alongthe 5.5 m long wire is ± 1.6 mm.

The SAMUS chambers were tested [76] in a 40GeV/c muon beam. The time-distance relation wasmeasured using beam proportional wire chambers . Theresolution was determined by directing a test muonbeam along three contiguous SAMUS tubes and mea-

I I

L1

2

4

6

8 10 '2 14

DRIFT DISTANCE (rrrr)

Fig. 65 . Drift time to distance relation for the SAMUS drifttube .

EE

Zor=

ô

Q

00

0 .9

08

07

0 .E

os

c2

DO Collaboration /Nucl. Instr. and Meth. in Phys . Res . A 338 (1994) 185-253

CF4 (g07) + CH, (10?~ )

0 2 4 e 8 ic ^2

DRIFT DISTANCE (mm)

Fig. 66 . Coordinate resolution of the SAMUS tubes vs driftdistance .

suring the deviation of the middle tube drift distancefrom the average of the first and last distance . Fig . 65shows the time to distance relation and coordinateresolution vs drift distance . Averaged over tube radius,the resolution of a single drift tube is 0.35 mm, as

LEVEL 1TRIGGERCONTROLCOMPUTER

PROGRAMMING. CONTROLAND MONITORING INFORMATION

HOST

shown in Fig . 66 . The operating anode potential of theSAMUS drift tubes is 4.00 kV.

Studies of the gas properties for the muon cham-bers have been made [74,75,79] . For the gases includ-ing CF4 , the drift velocity variation was shown to beabout 3% for 10% change in electric field . The driftvelocity is sensitive to CF, content ; this gas content isspecified with a precision of ±0.1% .

5 . Trigger and data acquisition

The DO trigger and data acquisition systems areused to select and record interesting physics and cali-bration events. The trigger has three levels of increas-ingly sophisticated event characterization . The Level 0scintillator-based trigger indicates the occurrence of aninelastic collision . At a luminosity of Y= 5 x 10 30cm -2 s -1 the Level 0 rate is about 150 kHz. Level 1 isa collection of hardware trigger elements arranged in aflexible software-driven architecture which allows easymodification . Many Level 1 triggers operate within the3.5 ws time interval between beam crossings and thuscontribute no deadtime . Others, however require sev-eral bunch crossing times to complete and are referredto as the Level 1 .5 triggers . The rate of successfulLevel 1 triggers is about 200 Hz ; after action by theLevel 1 .5 triggers, the rate is reduced to under 100 Hz .

EVENTS THAT PASS LEVEL 2TRANSFERED TO THE HOST

IFig . 67. Block diagram of the trigger and data acquisition system .

SUPERVISORENABLES A

LEVEL 2 NODETO RECEIVETHE EVENT

223

18 ENT TRICKIER NUMBERMUON MUON MU014 FOUND START DIGMATION SIGNALS FRONT-END FRONT-END

CRATE DATEDETECTOR TRIGGER ABOVE Pt THRESHOLD FRONT-END BUSY SGNALS CRATES

z PosTIGN.LEVEL ZERO LEVEL ZERO aEAM-GAS.

TWO INTERACTIONS FIRST LEVEL TRIGGER DATA BLOCKDETECTOR TRIGGER LEVEL 1

ANDLEVEL 1.5TRIGGER DATA CABLE DATA CABLE

TRIG I T T~ TRD CONFIRMATION OF FRAMEWORKS SEQUENCERS BUFFER-DRIVERS

DETECTOR TRIGGER ELECTRONCANDIDATES SPECIFIC THE SEQUENCERS

TRIGGERS SELECT THE CRATES EVENTSABLE OF FIRED TO BE READ ON REMOOT

ONELECTRONCANODATES

EACH DATA CABLE DATA CABLESS'S12810 EM STROBE

TPoGOER TOWERS LEVEL 1 GLOBAL ENERGY ANDMOMENTUM SUNS.CALORIMETER CALORIMETER - LEVEL 2 LEVEL 2

1200 HD TRIGGER LOCAL ENERGYCLUSTER COUNTS SUPERVISOR PROCEEMSOR

TRIGGER TOWERS FAR

224

Candidates from Level 1 (and 1.5) are passed on thestandard DO data acquisition pathways to a farm ofmicroprocessors which serve as event builders as wellas the Level 2 trigger systems. Sophisticated algorithmsreside in the Level 2 processors which reduce the rateto about 2 Hz before passing them on to the hostcomputer for event monitoring and recording on per-manent storage media. A block diagram of the triggerand data acquisition system is shown in Fig. 67 .

5.1 . Level 0 trigger

The Level 0 trigger registers the presence of inelas-tic collisions and serves as the luminosity monitor for

Do Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

J11 1"111111111

=HIM04

the experiment . It uses two hodoscopes of scintillationcounters mounted on the front surfaces of the endcalorimeters . These hodoscopes have a checker-board-like pattern of scintillators inscribed within a 45 cmradius circle giving partial coverage for the rapidityrange 1.9 < ,r7 < 4.3 and nearly full coverage for 2.3 < 77< 3.9 . The rapidity coverage is set by the requirementthat a coincidence of both Level 0 detectors be >_ 99%efficient in detecting non-diffractive inelastic collisions .

Fig. 68 shows the layout of the Level 0 counters .Two planes rotated by 90° are mounted on each end-cap calorimeter 140 cm from the center of the detec-tor. Each hodoscope has 20 short (7 X 7 cm2 squares)scintillation elements with single photomultiplier read-

Fig. 68 . (a) One plane of the Level 0 trigger hodoscopes. Active scintillator is shown shaded . (b) Two planes rotated by 90°constitute a complete Level 0 hodoscope.

20a

wis

10

5

0ô

FISN Short Counters

Long Counters

DO Collaboration /Nucl. Instr. and Meth. in Phys. Res . A 338 (1994) 185-253

50

i00

150

200

Time Resolution (ps)

Fig. 69. Time resolution distribution for the Level 0 countersfrom cosmic ray tests .

out and eight long (7 x 65 cm2 rectangles) elementswith photomultiplier readout on each end. To obtaingood timing characteristics, 1.6 cm thick Bicron BC-408PVT scintillator and 5 cm diameter Phillips XP-2282photomultiplier tubes are used. Light pulses from aUV laser are distributed to each photomultiplier tubevia fiber optics for use in calibration and monitoringthe detector .

In addition to identifying inelastic collisions, theLevel 0 trigger provides information on the z-coordi-nate of the primary collision vertex. The large spreadof the Tevatron vertex distribution (Q= 30 cm) has the

Detector Platform

potential for introducing substantial error in the ETvalues used in the trigger. To provide ET corrections, avertex position resolution of 8 cm is needed at Level 1and 3 cm at Level 2. The z-coordinate is determinedfrom the difference in arrival time for particles hittingthe two Level 0 detectors . The Level 0 counters pro-vide excellent time resolution since the active scintilla-tor elements are either small or viewed by photomulti-plier tubes on both ends so that time averaging can beused . The time resolution of each Level 0 counter wasmeasured in a cosmic ray test stand; Fig. 69 shows timeresolutions of typically 100-150 ps, which is wellmatched to the required vertex position accuracy .

At high luminosity, there is appreciable probabilityfor multiple interactions ; at Y= 5 x 10 30 cm -2 s-1there are on average 0.75 interactions per crossing. Inthe case of multiple interactions in one crossing, theLevel 0 time difference information is ambiguous and aflag is set to identify these events to the subsequenttrigger levels .A block diagram of the Level 0 electronics is shown

in Fig. 70 . PMT signals are amplified and split into tworeadout paths. One path makes an analog sum of thesmall counter signals for each hodoscope and uses aGaAs-based digital TDC to make a fast measurementof the vertex position for Level 1 ET corrections [80] .A cut I zv,x I < 100 cm is made to separate beam-beaminteractions from beam-gas and beam-halo events .The second path digitizes the time and integratedcharge for each counter. A more accurate slower mea-sure of the vertex position is made by applying fullcalibration and charge slewing corrections to the dataand using the mean time for each hodoscope to deter-mine the vertex position . The rms deviation of the timedifference is also computed and used to flag events

MCH

TokenRing

TriggerData Cable

slow Z . MITo Trigger

Fast ZTo Trigger

225

Fig. 70 . Block diagram of the Level 0 electronics . QTAC is a charge and time to amplitude converter; CFD is a constant fractiondiscriminator ; LOCTRL is a VME controller ; and VI is the DO vertical interconnect described in Section 6.7. The slow z

determination also produces a flag for multiple interactions (MI).

226

with multiple interactions . All Level 0 calculations areperformed in hardware and are available for use in theLevel 1 and Level 2 trigger decisions.

ADC

E

EMTRIGGERTOWERSIGNAL

1280EM-

THE LEVEL tCALORIMETER

TRIGGERSERVICESTRIGGERTOWERS

0 s IETAI (4 .4WITH FULLAZIMUTHALCOVERAGE

Dß Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

TOTAL ET REFERENCE SET 1

TOTAL E REFERENCE SET 2

TOTAL ET REFERENCE SET 3

DOUBLEBUFFERED

MEMORY ARRAY

COUNTER TREE CO).mARATORSTOTAL ET REFERENCE SET 0

DOUBLEBUFFERED

MEMORY ARRAY

The Tevatron luminosity is obtained by measuringthe rate for non-diffractive inelastic collisions . Eventsof this type are selected by requiring a Level 0 coinci-

8 BIT DIGITAL EM TRIGGER TOWER ENERGY SIGNALREAD OUT IN THE FIRST LEVEL TRIGGER DATA BLOCK

SCALED 8 BIT DIGITAL EM TRIGGERTOWER SIGNAL TO THE EM ADDER TREE

ADDER TREENAnaoNic Er /COe~-GLOBAL HADRONIC ET i THRESHOLDN

(0 ~ N ~ 31)

EM Et TRIGGERTOWER > THRESHOLDAND HD Et TRIGGERTOWER < THRESHOLDTO THE EM CLUSTER

COUNTER TREE

TOTAL Et TRIGGERTOWER > THRESHOLDTO THE TOTAL Et

CLUSTER COUNTER TREE

SCALED 8 BIT DIGITAL HD TRIGGERTOWER SIGNAL TO THE HD ADDER TREE

8 BIT DIGITAL HD TRIGGER TOWER ENERGY SIGNALREAD OUT IN THE FIRST LEVEL TRIGGER DATA BLOCK

GLOBAL TOTAL ET i THRESHOLDN(0~Ni31)

MISSING ET

compARATORs

MISSING ET i THRESHOLD(O1N131)

COUNT OF TOWERS WITH EM ET ABOVE REFERENCE SET 0 :THRESHOLDN

COUNT OF TOWERS WITH EM ET ABOVE REFERENCE SET 1 aTHRESHOLDN

COUNT OF TOWERS WITH EM ET ABOVE REFERENCE SET 2 xTHRESHOLDN

COUNT OF TOWERS WITH EM ET ABOVE REFERENCE SET 31THRESHOLDN(01N131)

COUNT OF TOWERS WITH TOTAL ET ABOVE REFERENCE SET 0 1THRESHOLDN

COUNT OF TOWERS WITH TOTAL ET ABOVE REFERENCE SET 1 :THRESHOLD

COUNT OF TOWERS WITH TOTAL ET ABOVE REFERENCE SET 2sTHRESHOLDN

COUNT OF TOWERS WITH TOTAL ET ABOVE REFERENCE SET 31 THRESHOLON(0-N131)

Fig . 71 . (a) Block diagram of the front end card of the Level 1 calorimeter trigger. (b) Detail of adder and counter trees used to giveET, ~T and multiplicity outputs .

TRIGGERTOWERS

4

EACH OFTHE 320

CTFE CARDSSERVICESEM AND

4 HADRONIC1280 TRIGGER

HADRONIC- TOWERSTRIGGERTOWERS

5.2. Level 1 framework

DO Collaboration /Nucl. Instr. and Meth. in

deuce with

I z�, I < 100 cm . Scalers count variousquantities : live crossings, coincidences satisfying thevertex cut, and single hits in groups of similar counterswith and without valid coincidences . These scalers al-low the luminosity to be measured independently foreach beam bunch and provide feedback to acceleratoroperations .

Operation of the Tevatron collider with six bunchesof protons and antiprotons gives 3.5 Vs between cross-ings . Any rejection of events which can be accom-plished in this time incurs no deadtime penalty. Thehardware calorimeter trigger and a part of the muontrigger satisfy this constraint, but the remaining portionof the muon trigger takes several bunch crossing timesto complete and is logically incorporated as a veto onevent transmission . A trigger based upon TRD energydeposits which has been designed but not yet imple-mented would also operate in this extended time inter-val . The overall control of these Level 1 trigger compo-nents and the interface to the next higher level triggerresides in the Level 1 framework [81,82] and is shownin Fig . 67.

The framework gathers digital information fromeach of the specific Level 1 trigger devices and chooseswhether a particular event is to be kept for furtherexamination . In addition, it coordinates various vetoswhich can inhibit triggers, provides the prescaling oftriggers too copious to pass on without rate reduction,correlates the trigger and readout functions, managesthe communication tasks with the front-end electronicsand with the trigger control computer (TCC), andprovides a large number of scalers which allow ac-counting of trigger rates and deadtimes .

The selection of triggers is performed with a two-di-mensional AND-OR network. The 256 latched bitscalled AND-OR input terms form one set of inputs tothe AND-OR network ; these bits bear specific piecesof detector information (e .g ., two calorimeter clustersover 5 GeV) . The 32 orthogonal AND-OR lines arethe outputs from the AND-OR network and corre-spond to 32 specific Level 1 triggers . Each of thesespecific triggers is defined by a pattern indicating, forevery AND-OR input term, whether that term is re-quired to be asserted, required to be negated, or is tobe ignored . Satisfaction of one or more specific triggerswhile free from front-end busy restrictions and othervetos results in a request for readout by the dataacquisition hardware . If Level 1 .5 confirmation of aspecific trigger is required, the framework forms theLevel 1 .5 decision and communicates the results to thedata acquisition hardware . The Level 1 frameworkassembles a block of information summarizing all theconditions leading to a positive Level 1 decision (and

Phys . Res. A 338 (1994) 185-253

227

Level 1 .5 confirmation if required) for transmission tothe succeeding levels of analysis .

Interactions with the Level 1 trigger system occurthrough the TCC. The configurations of the activespecific triggers are downloaded from the host com-puter to the TCC. The large tables of information usedfor programming and verification of the hardwarememories in specific Level 1 trigger systems are down-loaded to the TCC and stored on its local disk. Scalersand registers can be accessed through the TCC inorder to perform trigger system programming, diagnos-tics, and monitoring . The TCC software is based on theDEC VAXELN multitasking real-time system soft-ware . The TCC software uses a low-level hardwaredatabase which establishes aliases and descriptions ofspecific hardware elements allowing easy reference,and a high-level object-oriented database in which trig-ger configurations are defined and recorded . The trig-ger system is monitored continuously to verify that itsconfiguration files remain valid ; failures cause alarmsto be set in the general DO alarm monitoring system .The Level 1 trigger data block is passed along to thedata logging stream so that subsequent processors canrecompute the input information and thus confirm theLevel 1 decision .

5.3. Level 1 calorimeter trigger

A schematic diagram of the logic of the Level 1calorimeter trigger [81,82] is shown in Fig. 71 . Thesystem operates on the analog trigger pickoffs from thecalorimeter BLS circuits, summed into A77 = i10 = 0.2trigger towers out to 1771 = 4.0. Separate trigger inputsare provided for the EM and fine hadronic sections(1280 of each) of the calorimeters .

Each input signal voltage is analog-weighted by thesine of the trigger tower polar angle to give the trans-verse energy appropriate for an interaction vertex atz = 0 . This weighted signal is then digitized in a fast8-bit flash ADC (20 ns from input to output) andclocked into latches, allowing pipeline synchronizationof all calorimeter information . The latches can also besupplied by test information allowing diagnostic studiesof all subsequent trigger functions . This pipelined ar-chitecture is compatible with future Tevatron upgradesincorporating shorter bunch times (and thus requiringa larger number of crossings for decision making).

The 8-bit digital information from the flash ADCprovides part of the address for several lookup memo-ries. The Level 0 trigger provides an additional 3address bits indicating the interaction z-vertex . Thelookup memories provide both EM and hadronic trans-verse energies for each trigger tower above a fixed cutbased on both electronics noise and physics considera-tions, and corrected for the z-vertex position, if known .The sum of EM and hadronic transverse energies for

22 8

DoCollaboration INucl. Instr. and Meth . in

each trigger tower is formed and stored in a 9-bitregister as an input for future, more powerful, hard-ware triggers . This sum is used to form the x- andy-components of transverse energy using lookup mem-ories . Lookup memories also return the EM andhadronic transverse energies for each trigger tower,without the cut and z-correction . The global sums ofthe six energy variables returned from the lookupmemories for each trigger tower are computed over thefull set of trigger towers by pipelined adder trees . Theadder trees are arranged such that geographically con-tiguous regions of the calorimeter are kept together.This permits the creation of intermediate energy clus-ters over larger regions of the detector, which may beused by future, more sophisticated, cluster handlinghardware .

The full event missing transverse energy, 9T, isformed from the x- and y-component global transverseenergies . The global total transverse energies, bothcorrected and uncorrected, are formed from the cor-rected and uncorrected global EM and hadronic trans-verse energies respectively . These seven energy vari-ables (global corrected EM ET, global correctedhadronic ET , global corrected total ET, missing ET,global uncorrected EM ET , global uncorrectedhadronic ET, and global uncorrected total ET) areeach compared with up to 32 programmable thresh-olds . Each such comparison provides a Level 1 frame-work AND-OR input term .

In addition to the sums described above, each indi-vidual EM trigger tower transverse energy is comparedto four programmable reference values . The corre-sponding hadronic trigger tower transverse energy islikewise compared to four different programmable ref-erence values . For each trigger tower, a bit is set foreach of the EM reference values exceeded, if notvetoed by the corresponding hadronic trigger towerenergy exceeding its reference value. The total trans-verse energy (sum of EM and hadronic transverseenergies) for each trigger tower is also compared to aset of four programmable reference values, producingfour more bits . These 12 reference values are sepa-rately programmable for each individual trigger tower.For each of these four EM ET and four total E,rreference value sets, the global count of the number oftrigger towers with transverse energies exceeding theirreference values is computed by summation over thefull set of trigger towers in the pipelined counter trees .The global counts for each reference set are thencompared with up to 32 programmable count thresh-olds ; these comparisons provide Level 1 frameworkAND-OR input terms.

5.4. Level 1 muon triggers

A block diagram of the muon trigger system isshown in Fig. 72 [83] . The basic information provided

Phys . Res. A 338 (1994) 185-253

Fig . 72 . Block diagram of the main elements of the muontrigger system . The inputs from the muon chambers arelatches derived from bits set by a cathode pad hit in a specific

muon proportional drift tube .

by the wide and small angle muon chambers to themuon trigger system is a single latch bit for each of theapproximately 16 700 drift cells of the muon system .This information gives the bend coordinate of hit driftcells with a granularity of 10 cm in WAMUS and 3 cmin SAMUS. These bits are received in the MCH by 200module address cards (MACS) which reside in 24 VMEdigitizing crates using custom backplanes . Each muondigitizing crate has a 68020 microprocessor used fordownloading data and event building . The MAC cardsand subsequent Level l and Level 1 .5 muon triggerelectronics are kept physically distinct for the fiveseparate eta regions of the muon detector (CF, EF-North, EF-South, SAMUS-North, and SAMUS-South).

The MAC cards perform zero-suppression and gen-erate bit patterns corresponding to hit centroids for theLevel 1 and Level 1.5 muon trigger electronics . Acentroid is defined as the most likely half-cell (5 cm inWAMUS and 1.5 em in SAMUS) traversed by a track,projected to the midplane of a chamber. Centroid PALlogic is programmed using pairs of drift cells and canfind the correct centroid even in the presence of geo-metrical inefficiencies or delta rays . The MAC cardsproduce a bit pattern corresponding to a logical OR ofthe centroids for use by the Level 1 muon trigger(called the CCT or coarse centroid trigger). In WA-MUS (SAMUS) this OR is performed on a group of

DO Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

three (four) centroids . The MAC cards also produce afull list of centroids to be sent to the Level 1 .5 muontrigger (called the OTC or octant trigger card).

The WAMUS CCT cards further OR the centroidbit pattern by a factor of four to give a hodoscopicpattern six cells wide (60 cm). Each CCT can acceptinputs from up to 13 MAC cards . In the CF region thiscorresponds to 3, 5, and 5 inputs from the A, B, and Clayers . The B and C layer MAC bit patterns are inputto two PALs which jointly produce a 12-bit outputpattern corresponding to 12 possible roads for A layerbits . Comparison with the actual 12-bit A layer patterndetermines good Level 1 trigger muons . Other etaregions such as SAMUS and overlap regions in whichtracks begin in SAMUS chambers and continue intosections of the WAMUS chambers also use CCT cardsfor Level 1 triggering . For example, in the SAMUSregion, CCTs first find x, y and u space-point tripletsin each of the three layers . Bits from good triplets arethen used to search for independent x, y and u roadsof 12 cm width . Finally, bits from the valid roads areused to find triple coincidences corresponding to goodA layer space points .

The output of all CCTs for a given eta region is sentto a second CCT-like card which counts muon candi-dates in that region . Two bits of muon multiplicityinformation per region are sent to the trigger monitorcard (TRGMON) described below . CCT decisions areavailable within the 3 .5 Vs bunch crossing time . Theresults from individual CCT cards are latched and readout using the CCT LATCH card in the OTC VMEcrates .

After a trigger framework decision on any Level 1muon trigger, the MAC full centroid lists are strobedinto the OTC cards for Level 1 .5 confirmation with asharper PT threshold . Each OTC accepts inputs fromthree layers of MACs. The OTC compares all combi-nations of A, B, and C layer centroids to determine ifthey correspond to tracks above a P T threshold (typi-cally 3-7 GeV/c) . The A, B, and C layer centroids areused as addresses to a pair of SRAMs whose patternscorrespond to tracks above a given PT threshold . A4 x 4 array of these SRAMs allows the lookups for the16 combinations of 1 A, 4 B and 4 C centroids to becarried out in parallel . Processing times for the Level1 .5 trigger in WAMUS regions are typically less than 2vs .

The SAMUS Level 1 .5 trigger faces a large combi-natoric problem due to the large flux of small-anglebeam-jet particles . In the SAMUS region, OTCs areused to first find separate roads in x and y for thethree layers and good x, y, u space-point triplets inthe B layer . Centroids from these OTCs are sent to asecond level of OTCs which link the B layer centroidsfrom x and y roads with the B layer centroids of goodx, y, u triplets . Combined processing times for both

stages of SAMUS OTCs can exceed 100 ~Ls; each OTCcontains a programmable timeout which aborts triggerprocessing in the case of very long processing times .

Further processing is done for each good OTCtrigger combination using the latched input centroidsas address inputs for a second set of memories . Thesememories produce two 24-bit user defined trigger wordswhich are placed in FIFOs for readout by the OTCmanager (OTCMGR) card . Presently these triggerwords are simply the centroids for good triggers . Aftereach OTC's processing is complete, its output FIFO isread by the OTCMGR. This information is availablefor further processing both on the OTCMGR and atLevel 2. A status word for each defined OTC is readout as well . The OTCMGR uses the centroid informa-tion contained in the trigger words to apply a secondP T threshold to the event, giving the flexibility of twodifferent transverse momentum thresholds . TheOTCMGR for each eta region processes all the triggerwords for that region and sends three user-defined bitsto the TRGMON corresponding to transverse momen-tum, multiplicity, and geographic information . Uponreceipt of good Level 1 or Level 1 .5 trigger from theframework, all trigger words in the OTCMGR are readout by the VME buffer/driver (VBD) in each OTCcrate .

The TRGMON (trigger monitor) card receives twobits of Level 1 muon multiplicity and three bits ofLevel 1 .5 information from each eta region . TheseLevel 1 and Level 1 .5 eta region bits are mapped into16 Level 1 and 16 Level 1 .5 physics bits (e .g . twomuons anywhere in 1711 < 3.4) via downloadable RAMon the TRGMON. These 32 bits are sent to the AND-OR network of the trigger framework for full triggerformation .

5.5. Level 1 TRD trigger

229

The transition radiation detector collects an excessof charge from converted X-rays produced by particlespassing through a stack of thin foils . The zero-suppress(ZSP) circuit following a flash ADC for each anodewire in the TRD contains a stream of signals withinsuccessive 10 ns buckets . The ZSP circuit sums thesesignals over time and makes them available for arefined Level 1 .5 TRD trigger, designed but not yetimplemented . The purpose of this trigger is to certifyelectron candidates found in the Level 1 calorimetertrigger . The TRD sums are joined in the Level 1 .5trigger [84] by a 5-bit representation of the interactionz-coordinate as determined from the Level 0 triggerand two 6-bit representations of the ?7 and 0 coordi-nates from the Level 1 EM calorimeter trigger towers .The calorimeter tower 77 -0 coordinates are strobedinto FIFOs in the order of decreasing likelihood for

23 0

being an electron . Each A0 = 0.2 calorimeter tower ismatched to eight TRD anode wires.

5.6. Data acquisition architecture

Do Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

The DO data acquisition system and the Level 2trigger hardware are closely intertwined and are dis-cussed together in this section. The system [85-901 isbased upon a farm of 50 parallel nodes connected tothe detector electronics and triggered by a set of eight32-bit wide high-speed data cables . The nodes consistof a VAXstation processor coupled via a VME busadaptor to multiport memory (for receiving data), anoutput memory board and an option for an attachedcoprocessor . The function of the Level 2 system is tocollect the digitized data from all relevant detectorelements and trigger blocks for events that successfullypass the Level 1 triggers and to apply those softwarealgorithms on the data needed to reduce the rate fromthe approximately 100 Hz input to about 2 Hz outputto the host computer and data logger . All the data fora specific event is sent over parallel paths to memorymodules in a specific, selected node (one of the 50).The event data is collected and formatted in final formin the node, and the Level 2 filter algorithms areexecuted . An overall block diagram of the Level 2system is shown in Fig. 73 .

Fully digitized data appears in the output buffers ofthe approximately 80 VME crates containing thecalorimeter and muon chamber ADCs and the trackingand TRD chamber FÀDCs about 1 ms after receipt ofa valid Level 1 or 1.5 trigger. In addition, the Level 0and Level 1 triggers themselves prepare registers con-taining information characterizing the trigger decision .Each crate of primary digitized data contains a 512kByte memory module with two data buffers . EachVME buffer driver board [91] (VBD) contains listprocessors for controlled data transfer from VME loca-tions in the crate onto an output data cable highway.Internal crate transfers occur at up to 30 MByte/s. TheVBD outputs for each particular sector of the detectorare connected sequentially to a high speed data cable.The data cable consists of 32 twisted pair lines for dataand 13 pairs for control and parity ; the clock rate onthe data cable is 100 Its for a data transfer rate of 40MByte/s/cable. Eight such data cables are providedfor the eight detector sectors : VTX, TRD, CDC, FDC,the two halves of the calorimeter (north and south),muon chambers, and Levels 0 and 1 triggers. The datacable segment lengths are restricted to less than 12 m;since most DO data cables must circulate over a longerpath than this, repeaters are provided to de-skew thesignals . Readout control for the VBD and arbitrationis achieved with a token passing scheme . When a tokenis received, the VBD external port processor comparesthe token bits with the crate buffers; if a match occurs,

DO DETECTOR

Fig . 73 . Block diagram of the Level 2 trigger and data acquisi-tion system .

pending buffers are transferred to the data cable. To-ken passes are clocked at 1 MHz.

The real time operation of the data acquisitionsystem is under the control of the Level 2 supervisorprocessor. A sequencer processor controls the opera-tion of the data cables through a set of sequencercontrol boards (one for each data cable). For eachvalid hardware trigger, the Level 1 system interruptsthe supervisor with the 32-bit pattern of specific trig-gers for that event, together with a 16-bit event num-ber. The superviser sends a 32-bit disabled-trigger pat-tern to the Level 1 trigger control computer thatchanges with time, depending on the state of the runor the availability of Level 2 nodes for specific triggerbits . On receipt of a Level 1 trigger, the supervisorassigns a Level 2 node and then interrupts the se-quencer processor. The sequencer hardware in turnfashions readout tokens for a specific list of cratesappropriate to the specific trigger pattern, and includeslow order bits of the event number to insure readoutintegrity . Token circulation and data readout on eachdata cable are managed in parallel by separate se-quencer boards . Provision is made for readout of anycombination of data cables collecting data from any

READOUT READOUT LEVEL-1SECTION " " SECTION TRIGGER

1

1

I Data Cable# I -

' Data .C;able#7'

v

e#S

i.

Do Collaboration INucl. Instr. and Meth. in Phys . Res. A 338 (1994) 185-253

VAXstation 4000/60

V ME-Crate

Output cable fromprevious Level-2 node

Data-cables from previous Level-2 node

Fig . 74.'Schematic diagram of each of the Level 2 nodes andassociated VME electronics . The VME crate contains theVME bus adapter (pVBA), a co-processor slot, four dualmultiport memories (MPMs) and the VME buffer driver for

output events (VBD2).

desired set of crates. Debugging and calibration opera-tions are handled flexibly in this way .

The eight data cables also circulate to each of the50 Level 2 processor nodes . Whereas the VME digiti-zation crates and the Level 1 triggers are in the MCH,the nodes are located in the fixed counting area . Thedata cable connections between MCH and the count-ing area are made through an optical isolator circuit tomaintain isolation of grounds between detector andfixed counting areas . Each Level 2 node, shownschematically in Fig . 74, consists of memory moduleswhich receive data from the data cable, a VAXstation4000-60 processor, an optional special function co-processor and an output VBD which buffers data fortransfer to the host computer . Central to each node isthe multi-port memory [91] (MPM) accessed by theprocessor, the input data paths, the optional coproces-sor, and the output VBD. Four MPM modules, eachhaving two channels of 2 Mbyte multiported memory,serve incoming data from each of the eight data cablesarriving at an aggregate rate of 320 Mbyte/s. TheMPM memory appears as contiguous 1/O space mem-ory to the processor . The incoming data is mappeddirectly into the desired raw data ZEBRA structures asdescribed in Section 7.2 . No copy operations are re-quired on the data in the Level 2 nodes and no

reformatting is necessary for subsequent offline analy-sis .

During the setup of a data acquisition cycle, thesupervisor polls the processor nodes to find one avail-able to be the target for the detector data . Particularevent types may be steered to specific nodes for specialcalibration purposes . Special coprocessors can be at-tached to the special function port of the MPMs toallow certain simple repetitive computations to be per-formed on the data rapidly, in parallel with the VAXprocessor analysis . Each node normally contains anidentical copy of the filter software which operatesunder the VAXELN system, although special calibra-tion or test nodes may be loaded with different code .Development of the software algorithms and controlcode is developed in high level languages (FORTRAN,C, and EPASCAL) and downloaded over Ethernet intothe nodes. A special "surveyor" processor monitors theoperation of the data flow, the processor node opera-tions, and the sequencer and supervisor nodes . Thesurveyor collects statistics for on-line monitoring of thefull system and provides diagnostics which are availablefor real-time displays and alarms . Once the Level 2filter code has selected an event for subsequent datalogging and analysis, the event and added informationfrom the Level 2 analysis are sent directly from theMPMs through a VBD to the host computer . Transferof the event from a Level 2 node to the host ismanaged in a similar way to the readout from thedigitizing crates into the Level 2 : a control processor("sanitizer") circulates a token to the Level 2 outputVBDs . A VBD with data dumps its event on theoutput data cable where it is received by a MPMmodule in a crate connected to the host system via aVME/XMI bus adaptor .

5.7. Level 2 filter

23 1

Software event-filtering in the 50 Level 2 nodes isexpected to reduce the approximately 100 Hz of inputrate to 2 Hz to be logged for offline analysis . Modelling[92] shows that to keep the deadtime due to Level 2processing below 2%, average processing time shouldbe < 70% of the average time between events sent to aparticular node, or 350 ms . The present processingtime is less than 200 ms, which comfortably meets thisbudget .

The VÀXELN filtering process in each node is builtaround a series of filter tools . Each tool has a specificfunction related to identification of a type of particleor event characteristic. Among the tools are those forjets, muons, calorimeter EM clusters, track associationwith calorimeter clusters, IET , and missing ET . Othertools recognize specific noise or background condi-tions . The tools are associated in particular combina-tions and orders into "scripts" ; a specific script is

232

3 VAXstation 4000-60

4 VAXstation 3100-76

12 X-Window Terminals

Do Collaboration /Nucl. Instr. and Meth. in Phys. Res . A 338 (1994) 185-253

Fig . 75 . DO host system computer configuration .

associated with each of the 32 Level 1 trigger bits . Thescript can spawn several Level 2 filters from a givenLevel 1 trigger bit ; for example, a single electrontrigger from Level 1 might be given different Level 2bits depending upon the ET threshold or other fea-tures present in the event (e .g . energy isolation or the

Filter

Processors

DataLoggingTask

AlarmServer

OtherServers J

Disk

S

presence of missing ET) . There are 128 Level 2 filterbits available in all .

Tools are typically developed offline in a VMSenvironment where they can be subjected to a varietyof tests and verifications using data or Monte Carloevents . Incorporation of a new tool or assembly of a setof scripts into the package for the Level 2 nodes isachieved with a facility called L2STATE [931 . TheL2STATE output consists of a set of FORTRAN filescomprising the tools, configuration files that specifythe arrangements of the Level 2 (and Level 1) filtertools, the specific parameters for use in the tools (e .g .thresholds) and lists of required resources. The COORtask of the on-line host computer program (c .f. section5.8) assembles this output and coordinates the down-loading of code, configuration files and resource lists tothe relevant lower level hardware processors .

The Level 2 nodes are coordinated through the hostcomputer . provision is made for run-time distributionsof parameters to the nodes, for collecting statistics onthe processing history for the nodes, and collection oferror statistics and alarms .

HardwareDatabase

AlarmDisplayTasks

Fig. 76 . DO host system software organization .J*

Tape

Level 2Supervisor

Global SharedCommonServer

Eventa Display Dis avr Taskse J

Coordinating C Event0 F> Is a

Task MonitoringrTriggerControl (COOR)

m TasksComputer m

0

Token Ring Event CollectionCommunication Control ls a

Tasks Tasks

Run/TapeHardware Database

Token Ring MonitoringGateways Tasks High Voltage

Control H is aTasks

Do Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

Filter tools are typically developed using FOR-TRAN. Incorporation of these tools into the ELNenvironment imposes certain constraints . StandardFORTRAN-77 will run in ELN, but I/O and system-calling services require special handling . There is noprovision for virtual memory in the Level 2 operatingsystem, so all code and constants must fit in the avail-able 8 Mbyte memory . Operation of multinode systemshas occurred for most of the DO beam and cosmic raytests [94] as well as for the generation of Monte Carloevents in the farm system [95] .

Preparation of triggers and modelling the timingand rate-rejection properties is done using a pair oflinked utilities called LISIM and L2SIM . Respectively,these programs operate on simulated detector data,using simulated Level 1 trigger elements and theAND-OR network, or the actual configuration filesand filter scripts for Level 2 . A standard set of physicsand background events, generated by ISAJET andconverted to simulated hits in all detectors by

CalibrationControl Files

23 3

DOGEANT (see Section 7.3), is used for these simula-tions . The outputs of these calculations are used tooptimize both the Level 1 trigger specifications and theLevel 2 scripts within the overall constraints of therates and rejection factors .

5.8. Host computer and on-line software

The host cluster consists of VAX 6620, VAX 6410and VAX 8810 processors and a set of shared disksconnected by the high speed Digital Equipment CIcluster protocol, together with 12 VAXstations con-nected by an Ethernet/FDDI network. The 6620 is theprimary data collection engine receiving events fromthe Level 2 output data-cable . It logs these events to astaging disk and dispatches a sample to the variousworkstations for on-line monitoring purposes . Record-ing occurs at rates of up to approximately three 500kbyte events per second . The 6410 is the primarymachine responsible for spooling events from the stag-

Operator

Fig . 77 . Schematic diagram showing the operation of the calibration procedure .

23 4 DO Collaboration INucl. Instr. and Meth. in Phys . Res. A 338 (1994) 185-253

ing disk to 8 mm tape. It also performs downloadingoperations . The 8810 is devoted primarily to hardwaremonitoring, but also participates in downloading oper-ations .

The host cluster also serves as the primary humaninterface to the detector systems. In this role it isresponsible for high level control of the data-takingsystem, downloading of all settable parameters, specify-ing hardware monitoring activities, and the recordingand displaying of data produced by the detector . Thesedata include p-p interactions, calibrations, monitoringinformation and alarms .

The event data-taking system is designed to supportflexibly defined partitions of the detector, allowing formany users to collect individually tailored data streams.Each such user has his or her own runs, triggers, eventfiltering or processing nodes, recording streams andevent monitoring tasks.

Fig. 75 shows the hardware configuration of thehost cluster and Fig. 76 shows key processes of thesoftware system . These processes are distributed acrossvarious members of the cluster. Communication be-tween processes is accomplished by a package calledinter-task communication (ITC) developed by DO. ITCworks transparently between processes within a singlenode or across nodes of the cluster . Data samples aresent to monitoring, display or calibration processesrunning within the cluster over a DO developed net-work extension of the standard global shared common(GSC) system supplied by the Fermilab ComputingDivision . Events are inserted into GSC through a sub-process of the data logging task which receives theevents from the Level 2 filter nodes. Events are se-lected from the extended GSC by consumer tasks,based on those filter bits satisfied during the filteringprocesses.

Event monitoring tasks are based on a programframework called EXAMINE. EXAMINE is one ofthe frameworks supported by the program builder(Section 7.2), and thus the application packages thatoperate within it are portable to the other standardDO frameworks described in Section 7.2. Interactionwith EXAMINE is through the COMPACK menusfacility described in Section 7.2 . Utilities within EX-AMINE present the user with lists of currently avail-able triggers and inform the user of ends of runs .Under normal data-taking conditions, there is an EX-AMINE running for each of the major detector subsys-tems as well as global analysis and global display ver-sions .

The COOR task is responsible for the overall coor-dination of the data taking system and runs as adetached server process. Client tasks connect to COORto request specific configurations of the detector, or torequest operational changes. In turn, COOR has con-nections to various data-taking subsystems through

which it sends download and control messages . Thetargets for these messages include various parts of theLevel 1 and Level 1.5 trigger system, the data gatheringand event filtering components of the Level 2 system,the front-end digitizing crates, pulsers, various timingand gating logic, and the data logging system . Thepaths for these messages include FDDI, Ethernet andIBM token-ring networks .A typical use of the various elements of the host

cluster and related processors for the electronic cali-bration of the calorimeter electronics is shown in Fig.77 . Under operator control, the host computer sendsconfiguration files to Level 2 nodes and Level 1 trig-gers to start processing special electronics calibrationdata . It issues commands over the control data path topulsers situated in the various calorimeter preamplifierenclosures. Pulse pattern sequences are generatedwhich flow along the normal data collection pathsthrough digitization, Level 2 nodes, and into the shareddata space in the host . Computations involving thepulser data are done either using filter scripts in theLevel 2 nodes or in tasks operating on one of theelements of the host cluster . The associated input datamay be recorded, but in most cases only the resultantpedestals, gains, bad channels, etc. are retained .

6. Detector services

DO has chosen wherever possible to unify its sup-port systems across the detector subsystem boundaries .In this section, we briefly describe some of the mechan-ical, electrical and monitoring systems which serve thedetector as a whole.

6.1 . Support platform

The detector support platform [96] is a box-beamwelded structure consisting of four beams assembledinto a "tic-tac-toe" pattern. The beams parallel to theaccelerator beamline are 13 .4 m long ; the transversebeams are 11 .6 m long. The box cross-sections are 1 .3m high by 0.8 m wide . Track plates are mounted on thetops of the beams to allow motion of the muon toroids(the CF shells retract perpendicular to the beams andthe EF magnets move parallel to the beams) . Thecentral beam portion of the CF toroids is fixed to theplatform (see Fig. 52) directly under the Tevatronbeamline and serves as the support for the calorime-ters . The end calorimeters retract along this beam togive access to the innermost tracking detectors .

The space within the platform is used to houseelectronics, gas, power, cryogenic and cable utilities forthe detector . Access to this interior space is given byholes in the central sections of the box-beams parallelto beamline . The five non-corner cells of the "tic-tac-

6.2. Cable plant

Do Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

toe" pattern contain electronics racks which house thehigh voltage fanouts and current monitors and theshaping circuits for tracking, TRD and calorimeterdetectors . The corner cells are reserved for the incom-ing and outgoing detector services: ac power, driftchamber gas connections, calorimeter cryogens, andthe merge point for cable communication to the radia-tion safe areas.

The platform is the transporter for the detectortravel between the collision hall and assembly area .Twenty 500-ton Hillman rollers are mounted under thetransverse box beams to allow movement on hardenedsteel tracks in the building floor . Vertical lift for theplatform is provided by 500-ton hydraulic cylindersmounted on each of the rollers . Independent operationof these lift cylinders allows the detector to be levelledat beam height . The platform is moved on the rollersby two pairs of hydraulic push/pull cylinders attachedon one end to the ends of the transverse box beams.The other end is fitted to a pin which is dropped intoholes in the track plates . The hydraulic rams contractto pull the detector over the 24 in . interval betweenholes.

DO contains approximately 5.5 x 106 m of cablepairs for distribution of signals, high voltage, monitor-ing, and controls. The cable connections to the DOdetector must penetrate the 12-ft thick concrete blockshield wall which separates the collision hall from theradiation-free assembly hall where the digital electron-ics and control areas are located. The main cable pathsegments are (a) detector preamplifier to shaper/mul-tiplexer ; (b) shaper to digitizers ; and (c) digitizers toevent builder/Level 2 trigger.

Preamplifiers are located as near the physical detec-tor elements as possible . Each detector element signal(approximately 120 000) is carried by its dedicated ca-ble to the shaper electronics . Articulated cable traysare provided for the preamplifier to shaper cables fromdetectors which can be moved. The shaper crates aretypically located in the detector platform (muon cham-bers are the exception where shaping is done locally atthe chambers). The central detector signals are trans-ported on 73 fl 28 AWG coaxial cables assembled into18 wide ribbon pairs, terminated in 40-pin cable con-nectors. Calorimeter signals are transported on 110 fltwist and flat (TnF) 25-pair wide ribbons.

The shaper outputs, multiplexed in the case ofcalorimeter and muon detectors, are gathered at the"cable corner" of the platform and carried over a cablebridge through the shield wall and to the movingcounting house (MCH) which houses the digitizers .The cable bridge is an open channel, lined with coppersheet. It has a total length of about 8 m and is approxi-

23 5

mately 235 cm across and 50 cm deep . The bridge isattached at one end to the cable corner of the platformand on the other end to the MCH. It is hinged in themiddle and can be folded vertically to allow relativemotion between MCH and platform . With the detectorin the collision hall, the bridge is fully extended andfilled with bags of glass beads to fill the remaininginterstices, and covered with a copper lid to completethe shielding against noise pickup . In this position, thecables pass through a penetration through the stackedblock shield wall . On leaving the bridge, cables aredirected upward through a cable tower fixed to theside of the MCH. Bundles of cables in the bridge andtower are organized such that at each of the threelevels of the MCH appropriate cable streams are peeledoff to deliver signals to the relevant electronics circuits .The cable tower also houses the ultraviolet laser whichfeeds calibration light to the Level 0 and ICD detectorsalong fiber optic cables . Central detector shaper toMCH cables are ribbons of coax pairs similar to thoseused from preamplifier to shaper. Calorimeter multi-plexed signal and muon latch signals are transportedon ribbon TnF cables . Calorimeter trigger and muonchamber time and pad analog signals use the 110 fZribbon coax cables.

High voltage for all detectors is distributed frompower supplies to active elements on eight-conductorcable with a single outer ground sheath. Special eightpin connectors and plugs are provided to allow discon-nection of the high voltage path . Connectors are speci-fied to give less than 10 nA current when ±10 kV isapplied to odd and even pins . The full DO detector issupplied from approximately 1000 HV supplies de-scribed in section 6.4 . Successive HV fanouts to indi-vidual detector elements are located in both the MCHand the detector platform . In the case of the calorime-ter, where current draw information is desired forindividual HV interleaves, special current monitoringfanouts are provided .

All cables in the DO detector were provided with abar code for specific identification and were enteredinto the cable database . The database contains infor-mation about the manufacturer, connector and cabletypes, slot/crate/rack address for both ends, as well asinformation about the detector channel served by thecable with the appropriate channel address name andnumber . This database enables easy access to the col-lection of cables, connectors, crate, slots, etc. impli-cated in any specified detector channel.

6.3. Mooing counting house

The MCH is a three story enclosure housing theLevel 0 and 1 trigger electronics, high voltage powersupplies, and digitization electronics for all detectors .It supports the load of the cable bridge and moves on

236

Trip Bus

wheels along the side of the assembly hall . Two park-ing positions are provided with doorways from thefixed counting areas; one serves the MCH when thedetector is in the assembly hall and the other is usedfor the detector in the collision hall . The MCH iselectrically isolated from the building ground and iselectrically connected to the detector through the cablegrounds. Cables enter each level of the MCH from thecable tower into space beneath a removable tile com-

21

a.ka

OskA

CMUM.OdVdut.

Extemal Trip (Bi-directional)

Do Collaboration /Nucl. Instr. and Meth . in Phys. Res . A 338 (1994) 185-253

. . . . . . . . . . . . . . . . .Repedfhf 8 Slis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C . _____1

. .ateor eac oupp e locksA &B F

Fig. 78. Block diagram of the HV control board.

+ 12 volt

Fig. 79 . Block diagram of the HV dc-to-dc supply circuit .

puter floor . Each floor of the MCH contains 25 elec-tronics racks accessible from front and rear . Almost allcrates are VME standard . The lowest floor nearest thedetector houses the Level 0 and 1 trigger framework,calorimeter trigger, and HV power supplies for alldetectors . The middle floor contains the master detec-tor clock and the flash ADC electronics for the track-ing and TRD detectors, together with the MCH end ofthe Level 2 control and data cables . The top floor is

Do Collaboration I Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

devoted to the calorimeter and muon digitizers and themuon trigger elements .

6.4 . High voltage power

In order to provide compact high voltage power,standardize monitoring and computer control and sim-plify maintenence, DO has developed a new computercontrolled VME-based HV power supply [97-99] . Thesupply is a 6U x 60 mm wide unit containing eightindividual outputs . Six of these supplies are housed ineach crate, together with a crate monitoring moduleand the 68000-based HV control module . Three mod-ule types were produced . Two are negative and posi-tive polarity supplies operating between 10 V and 5 .6kV (in 1 .36 V increments) at < 1.0 mA. The third is apositive supply operating between 10 V and 2.0 kV (inincrements of 0.49 V) at <_ 3 .0 mA. Each supply unitcontains a control board which supports the VME-businterface, digital controls, and ADC/DAC functions .Eight self-contained plug-in units (pods) mounted onthe control board contain the dc-to-dc converter cir-cuits which supply the high voltage. Any mixture of HVpods can be installed on a single control board . Fig . 78shows a block diagram for the HV control board ; Fig .79 shows the dc-to-dc converter supply schematic . Eachpower supply unit is provided with a front panel SHVconnector and with a potentiometer for correction ofthe resistor tolerances .

The control board contains separate 16-bit DACsused for high voltage setting and 12-bit DACs forover-current trip setting, and a single 15-bit multi-plexed ADC for read-back monitoring of voltage, cur-rent, low voltage levels, temperature, and limit settingsfor current and voltage . HV trips are initiated byover-current or over-voltage using comparators on thecontrol board . Provision is made for grouping supplychannels together for common ramp and common tripin the event of any single member channel trip . Hard-ware trips can also be generated by external faultconditions (fire alarms, gas leaks, etc .). Upon receipt ofa trip, a latch is set to prevent turning on the supplywithout operator reset; the voltage DAC is also resetto 0 V . Front panel potentiometers provide an overridelimit to the maximum voltage available .

Each crate contains a control module with a 68020subordinate processor which is responsible for record-ing the recent history of the supplies in that crate, thusfacilitating information polling by higher level proces-sors . The subordinate processor also controls the indi-vidual supply conditions . A collection of crates is con-trolled by a master 68020 processor which governs thegroup trips and ramps . External communication withthe crate level processors is achieved over two distinctpaths . A PC-based host can be connected to the system

23 7

via a BIT3 interface or token ring ; this system is usedfor expert debugging and monitoring at the local (crateor rack) level . A VAX-based software system connectsto the HV power supplies through the standard DOcontrol path described below. It allows multiple userinteraction with the high voltage system and providesthe high level monitor and control required by thenon-expert experimenter . Facilities exist with eithersystem for logging supply conditions, producing his-tograms, strip charts, and parameter pages summariz-ing the system status .

Correction functions can be downloaded to the sup-plies to improve hardware inaccuracies . Measurementsshow that, with these corrections, the set voltages arewithin 0 .5 V of the nominal desired value . Currentreadback offsets are less than 32 nA. Output voltagetemperature coefficients are 300 ppm/°C .

6.5. Clock

Many separate processes for the operation of theDO detector require a precise train of timing pulses,synchronized with the accelerator clock. The DO clocksystem provides such pulses with a flexible controlsystem to allow user formulation of the clock pulsetrain desired for particular use .

The clock system hardware [100] consists of threemodule types : the phase coherent clock (PCC), se-quencer, and selector fanout . The PCC receives signalsfrom the Tevatron timing and rf sequences, and pro-vides synchronized signals at the first harmonic rf fre-quency of 106 MHz. In the event that the accelerator isnot operational, the necessary rf and Tevatron synchro-nization pulses can be generated internally in the PCC.The sequencer modules generate up to 23 independenttimelines of 1113 elements each, corresponding to the1113 rf buckets in one full turn of the Tevatron . Theselector fanout module receives the signals from thesequencer and distributes them to the various detectorsubsystems .

Multiple clients can be served by the DO clockinterface [101] for establishing one or more of the 11particular time sequences available for general users .The clock interface manages the various reserved time-lines and prevents interactions between the separateclient processes . Commands are available which allowopening, initialization and closing timeline requests,programming the specific pulse sequences desired, ref-erencing the sequence to a specific Tevatron time,establishing error handling response and requestingstate change notification messages for the user . Theselector fanout modules can also be programmed toreceive timelines via the COM TKR program on thehost VAX system .

23 8

6.6. Alignment

Do Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

Maintaining good knowledge of the locations of theDO subdetectors relative to each other and to theTevatron beams is essential . The DO alignment prob-lem is compounded by the fact that the detector is to agood approximation hermetic, so that inner detectorsare hidden from external view, and by the fact thatmajor pieces (CF, EFs, ECs) are moved for access andthus can return to different locations. In addition, thedetector is heavy and thus the floors of assembly andcollision halls themselves are deformable .

The detector has been fitted with several sensorsserving different purposes [102] . The liquid level sys-tem uses closed vessels of water connected by tubes ;the water level is sensed by capacitive proximity sen-sors . These devices [103] are mounted on the detectorplatform and tunnel magnets to provide vertical refer-ence over the full experiment to within about 100 wm .A tilt sensor system based on an electrolytic bubble tiltmeter [104] measures angular motions of the toroidelements to within about ±20 wrad . These measure-ments monitor tilts induced by magnetic forces andmovements of detector pieces . Proximity sensors [105]determine the distance to iron elements using eddycurrent detectors and allows positioning of the detec-tor or moveable subdetectors to a precision of about 25

wm.The history of changes in position and tilt sensors is

referenced to the values obtained at the time of thebaseline optical survey . The optical survey yields a fewmillimeter accuracy for absolute location of the largemuon chambers in the outer shell of the detector . Theinner tracking detector survey is somewhat more accu-rate . Final location of active detector elements relativeto the beam axis and to each other is determined byminimizing the fitted position residuals from particlesproduced in p-p collisions .

6 .7. Experiment monitoring and control

There are two independent pathways for communi-cation and data flow between the DO detector and thehost computer cluster . The collection of event informa-tion from the detector digitizer buffers is of primeimportance and is transferred over the high speed datacable highways discussed in Section 5.6 . Event dataacquisition is essentially a uni-directional process withthe minimum of controls . The bi-directional polling ofdetector hardware elements and front-end softwareprocesses, the collection of alarms, and the download-ing of the numerical specifications for digitizer anddata-acquisition crate operations uses a completely in-dependent control pathway.

6.7.1 . HardwareFig . 75 shows the block diagram of the hardware

monitoring system together with the event data pathand host system [106-108] . The on-line host computersystem described in Section 5.8 provides the high levelcontrol and user interfaces . An Ethernet LAN con-nects the host VAX processors via one of severalgateway processors to an IEEE 802.5, 4 MHz token-ringLAN. The gateways are microVAX processors whichare connected to both the Ethernet and token-ringLANs and which execute DO-developed software un-der the VAXELN operating system . The token-ringLAN services the front-end processors such as theMotorola 68020 processors located in the VME-basedelectronic crates in the MCH and in the high-voltagecontrol systems and the IBM-compatible processors inthe gas-flow control systems . No processors are in-stalled in the front-end analog signal processing cratesor on the detector platform to prevent noise genera-tion.

The 68020 processors control and collect informa-tion from remote readout modules called rack monitorinterfaces (RMIs) which are connected by a MIL-1553Bserial link and are distributed throughout the detectorelements, support platform electronics, and MCHracks . The RMIs collect information such as voltages,currents, temperatures, or water flows from all cratesin the experiment . Each RMI contains a 646-channeldifferential ADC, eight channels of DAC output, and32 bits of digital input and output. The MIL-1553 busserving the RMIs has no continuous clock and thus isquiet except at data transmission time . Although provi-sion was made to inhibit RMI data exchange duringevent acquisition times, this has not proved necessary .

The 68020 processors are also connected to thedigital electronics crates in the MCH by means of avertical interconnect card in a master processor cratewhich drives a single, 64-pair cable in a daisy chain of afew meters length to several slave crates . The verticalinterconnect allows for the downloading of the severalmegabytes of constants required by the DO trigger anddigitizing electronics without putting every digital crateon the serial network . The vertical interconnect allowsthe memory from five "slave" digitizing crates to bemapped to the master controller in a transparent way.Slave crates are allocated up to 2 Mbytes in the masterVMEbus address space . Data transfer is controlledwholly by the master crate to prevent bus contention.The vertical interconnect system allows for a variety ofdiagnostic and test modes of operation, as data can bedownloaded into the primary digitizers for each detec-tor and read back through the primary event acquisi-tion system . This allows a system check of the full dataacquisition system .

The control station processors provide the supervi-sion of the RMIs and vertical interconnect to the

digital crates. They acquire data at 15 Hz and buffer itfor transmission through the control data path to thehost system . Alarm scanning for hardware or softwareelements outside of preselected tolerances is done lo-cally in the controller . Only those data requested byhigh level processes are transmitted . Operation of thehardware monitoring system can be done indepen-dently of the central host VAX computer by means ofkeyboards attached to the local control stations . Thelocal stations utilize a local database in non-volatileRAM which is downloaded from the central database .

6.7.2 . Control softwareThe hardware monitoring system described in the

preceeding section is controlled from the on-line host

DO Collaboration /Nucl. Instr. and Meth.

UserDisplays

Fig . 80 . Dataflow diagram for the alarms system .

in Phys . Res . A 338 (1994) 185-253

239

computer (see Figs . 75 and 76), but is separated fromthe main event collection and physics monitoring path-ways . The processes in the host-level computers re-sponsible for control, hardware monitoring, and down-loading of run control and calibration data to thedetector electronics communicate over the control paththrough a library of standard interface procedurescalled CDAQ (Control Data AcQuisition) [109] .

At the heart of the monitoring function is therelational hardware database called HDB [110-113],based on the RdB [114] product . HDB defines all ofthe essential attributes of the hardware (rack, crate,card addresses, cable connections, applicable voltageor current settings, etc .) and provides CDAQ with theinformation it needs to establish communications with

SmartAlarm

Filtered

I ProcessAlarmMessage

'\ 1 .2

DeviceHierarchy

FE AlarmMessage

240 Do Collaboration /Nucl. Instr. and Meth . in Phys . Res. A 338 (1994) 185-253

a desired hardware device and to convey its description

to the front-end node in appropriate format for direct

access writes and reads. The HDB database is approxi-

mately 100 Mbytes in length . It contains information

on all detector systems, support systems such as low

voltage, high voltage power, cryogenics and argon pu-

rity monitoring, air handling systems, as well as soft-

ware systems such as CDAQ itself. Various facilities

for accessing the database have been provided, includ-

ing interactive and batch interfaces .

The procedures in the CDAQ library hide the de-

tails of the actual data-acquisition protocols by trans-

lating from a simplified user's view of a device with

multiple attributes to the required access path descrip-

tion via HDB queries. Requests in the form of multiple

device/attribute pairs, which may reference several

front-end nodes, are assembled into message packets

called frames destined for individual front-end nodes.

These requests are issued either in single-execution or

periodic mode . The CDAQ interface manages the syn-

chronization and assembly of replies from multiple

front-end nodes, arbitrates contention for shared re-

sources, and automates recovery of all allocated re-

sources, both local and network-wide, when a request

is deleted or the user's process terminates abnormally .

DO has developed a number of application pro-

grams which give dynamic graphics or text displays of

the experiment hardware status . Many use routines

from the CDAQ library to periodically gather informa-

tion from selected devices which are monitored or

otherwise accessible to the front-end computers. Typi-

cal examples of these programs are: (1) a general-pur-

pose, parameter page program which displays readings

of devices in numerical, strip-chart, or histogram for-

mat, (2) a high-voltage control program which provides

complete management and display functions for the

high-voltage hardware, and (3) a low-voltage program

which provides a similar service for the many low-volt-

age power supplies in the MCH and on the detector

platform .

6.7.3. Alarms softwareIn a complex collection of systems such as the DO

detector, there is a very large number of conditions

which can damage the quality of the data being col-

lected, or which threaten the detector systems them-

selves . These conditions can generally be sensed

through some measurable parameter (temperature,

power, voltage, current, humidity, etc.) going outside a

permissible range resulting in the generation of an

alarm . In an environment as strongly controlled by

overlapping software systems as DO, the continued

functioning of the software tasks must itself be moni-

tored. Since not all alarms can be examined immedi-

ately by the human operators, there is a need for

long-term alarm recording. Some attempt also needs to

be made to ascertain the common root cause for a set

of related alarms . The Do alarms package [115], writ-

ten mainly in PASCAL, has been created to handle

these needs. Fig. 80 shows the organization of the

alarms system software .The Do alarm system operates with a central "alarm

server" task on the host VAX cluster which receives

alarms from multiple sources and distributes alarms to

multiple clients. It may receive messages from front-end

processors which pass alarm information through thetoken-ring/Ethernet gateways . Other processes onVAX nodes reachable via DECNET may also generatealarms . Any DECNET accessible node may also be aclient .

The front-end processors monitor the hardware de-

vices at 15 Hz and generate alarms when the devices go

out of tolerance. Alarm messages contain the current

device reading together with the local copy of descrip-

tive information derived from the master HDB

database . Critical software processes also send periodic

"heartbeats" to the central alarm server. The absence

of a heartbeat, generally due to the disruption of the

process, causes the alarm server to generate an alarm

condition .There are two types of alarm messages . The first is

generated upon a state transition between an "accepta-

ble" and an "unacceptable" condition. The central

alarm server task maintains a list of devices in the

"unacceptable" condition which it can provide to client

tasks. The second type of alarm message, more appro-

priately called a "significant event", carries informa-

tion for distribution to other processes and is not

stored . For example, begin and end run notification

are propagated in this way.The central alarm server task collects alarms and

heartbeats from all sources. These are distributed to

client processes based upon filters which select classes

of events . Standard client processes include a process

to which to log events to a disk file, a process which

notes high priority alarms and is capable of interrupt-ing data acquisition, and a process for alarm display.

The interactive alarm display is built using the MO-

TIF windowing system . Customized displays can be

created to produce multiple lists of alarms which sat-

isfy different filters . In addition to the information

contained within the alarm message, detailed informa-tion obtained from the HDB database can also be

displayed for each alarm.

6.8 . Test beam system

Over the lifespan of a detector the size of DO,

there are many tests required for individual sub-detec-

tor systems as well as for larger collections of intercon-

nected systems. A general facility was developed in a

beamline of the Tevatron for this purpose . Many tests

of tracking chambers, muon chambers and calorimetermodules have been performed in beams of pions andelectrons ranging between 10 and 150 GeV/c. Re-cently, low energy particles were provided using ter-tiary beams [116] from targets located just upstream ofthe DO facility in the beam transport system. Bothelectrons and hadrons were delivered to the detectorswith momenta down to 1 GeV/c. A concrete shieldbehind the test facility allowed the use of tagged muons.

Several dedicated support systems were providedfor the tests. Proportional wire chambers with 1 mmwire spacing recorded the trajectory of beam particlesand allowed momentum determination with precision0.25% . Two gas-filled Cerenkov counters gave discrimi-nation of pions and electrons ; the electron contamina-tion of the pion sample selected was typically 10-4 . Aprecision table for supporting and positioning trackingchamber modules with remote translation and rotation

Do Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

Control ControlRequests Replies

was installed in the beam enclosure. A large liquidargon cryostat was built for testing a variety ofcalorimeter modules [57,117] . This cylindrical cryostatwas supported on a moveable platform allowing trans-lations in both horizontal and vertical directions androtations about the vertical and the cryostat axes .A trigger and data acquisition system was built

which was virtually identical to that in the final experi-ment, except that fewer elements were needed . Thusalthough providing a trigger for the test beam is farsimpler than in the collider, much valuable experiencein operating the Level 1 and Level 2 trigger systemsand hardware monitoring and downloading could begained prior to installation of the detector in thecollision hall . The host computer system was similar tothe final system and the on-line control and dataacquisition software was tested and evolved over thecourse of the test beam runs . A final benefit from the

24 1

Fig . 81 . Dataflow diagram for the Dß control path processes. The circular objects are subprocesses and the objects betweenhorizontal lines are data stores . Arrows indicate the directions of stimulii and data transfers.

242

simulation of the DO environment in the test beamfacility was the training of the DO physicists in operat-ing the detector and interpreting the on-line monitor-ing data .

7. Software

Do Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

A large body of software has been developed foracquiring data from the DO detector, for monitoringand controlling the hardware, for off-line reconstruc-tion, for Monte Carlo event simulation and for therelated databases. In this section, we summarize thesoftware and computational procedures developed withspecific application for the DO experiment.

7.1 . Structured analysis /structured design

It was recognized early in the design of DO that itwould be useful to adopt a more rigorous discipline forthe development of software than had been common inpast high energy physics experiments, since softwarewould be produced by a large number of widely dis-persed individuals with a moderately large turnoverover the lifetime of the project. There was a need forclear design specification and documentation of theproduced software . In addition, the broad dispersal ofthe effort placed a premium on clean delineation ofthe various sub-components of the project .

The methodology used for the software develop-ment

was

"structured

analysis/structured

design"

Fig . 82 . State transition diagram for high voltage control.

Enable -Request

(SA/SD) [118,119], developed for the commercial envi-ronment and previously adopted by the ALEPH Col-laboration at LEP [120] . SA/SD stresses two phases insoftware development . The first is an analysis phase inwhich careful specification of the software is preparedusing some well-defined graphical tools. The second isthe detailed implementation phase .A key element of SA/SD is the dataflow diagram

(DFD) which focusses upon the flow of data through asoftware system and the transformations which areperformed upon that data . The DFDs are arranged inhierarchies which describe succeedingly more re-stricted portions of the overall task. The stimuli anddata transfers from one subprocess to another areessential elements . The definitions of the data beingtransferred are specified in accompanying data dictio-naries . Particular examples of DFDs are shown in Figs .80 and 81 .

Working from the documentation produced in theform of the DFDs in the analysis phase, code is thendesigned during the second SA/SD phase which im-plements specific diagrams . The state transition dia-gram is introduced to represent the dynamic activationof processes and to focus on the conditions whichcause state transitions . An example state transitiondiagram is shown in Fig . 82 . The structure chart repre-sents the decomposition of tasks into functional mod-ules which are the units of implementation . It approxi-mates the traditional diagram of program logic andforms a primary component of the final software docu-mentation . Review and criticism of the code design wasconducted using these structure charts .

Our experience using SA/SD was good at the be-ginning of the project [121] . The initial work preparingand discussing dataflow diagrams and structure charts,though imposing some overhead, did enhance someclarity of purpose and agreement on the basic structureof the DO software system . The methodology alsoproduced very useful software documentation early inthe project . This documentation was not kept up todate over the long run, largely because no automatedtools existed to support the editing, integration andverification of the various diagrams and associateddictionaries .

7.2. Utilities and tools

DO Collaboration INucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

7.2.1. Data managementData management in DO is based on two products .

The first is RdB [114], a relational database used in thereal-time online environment . The second is ZEBRA[122] which is also incorporated into many of thegenerally available CERN packages . ZEBRA givesstandardized, hierarchical storage banks for commonlyused data, together with reference links to other re-lated banks . It is used to manage data of many differ-

243

ent types, including event data, detector descriptions(geometry, survey constants, etc .), calibration con-stants, run control parameters, graphics control param-eters and histograms . In the case of calibration con-stants and other data sets requiring complex access anadditional layer is placed over ZEBRA. This layer isDBL3 [123], developed by the L3 collaboration atCERN, which supplies a database set of utilities basedon ZEBRA.

The event data are put into ZEBRA format evenbefore on-line calculations are performed ; this is doneas part of the event building process in the Level 2filter nodes . The event data forms a single tree struc-ture and as reconstruction is done, extensive referencelinks among related bodies of data are established.

Various standard utilities are required to be sup-plied before any new ZEBRA bank is included in thelibrary . They include documentation files, a parameterfile giving the structural link, a bank booking routine, afunction returning the pointer to the bank, a routinewhich returns the values of the bank data words, and adump-printing routine .

Parameters controlling program behavior (such asvalues of cuts to be used in offline analyses) are kept ineditable ASCII "run control parameter" files . A set ofutilities have been developed which read these RCP[124] files, store the information in ZEBRA banks, getthe specific RCP parameter, or add information to anRCP file . A particular use of the RCP files is thestorage of detector geometry constants . Program use ofgeometry information is accomplished through the util-ities for interacting with the associated ZEBRA files .

7.2 .2. Code managementDO software is stored in a library maintained by a

set of procedures based on the CMS (code manage-ment system) and MMS (module management system)[114] . Some 45 distinct subdirectories exist in the li-brary, each associated with a specific software packagesuch as the DO Monte Carlo DOGEANT, the utilitiesfor ZEBRA manipulations, etc . Each such subdirec-tory is managed by a designated "czar" . The CMSlibrary keeps code sources, command procedures, doc-umentation, and in cases when the code is not main-tained by DO, object libraries .

There is only one central copy of the CMS librarywhich contains the update and release history of eachof the subdirectories, so that retreat to the softwarestatus of an earlier epoch is possible . Code developersare encouraged to create local CMS libraries on theirown clusters to facilitate local activity . At suitableintervals, updates are released by the library czar .Strict rules for reserving and replacing library modulesare enforced, so that all software developers know thestatus of changes and inconsistencies can be avoided .Release of code generally occurs in two stages. First, a

244 Do Collaboration INuel. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

test release is made to all nodes; users are notified andencouraged to test the new code extensively . In thisphase, test and prior official releases coexist in thelibrary . After the test period, the test release is de-clared official and replaces the official release.

The source and object libraries are automaticallydistributed to DO collaboration VAX computers sothat identical environments exist at all institutions .Release to other than VAX computers requires inter-mediate steps to convert source codes.

7.2 .3. Screen managementScreen management and the interaction of users

and DO software is accomplished with a commandinput package called COMPACK [125], written by DOin VAX FORTRAN and utilizing the VAX run timelibrary . This set of routines supports the input ofrun-time options, program specifications, program con-trol, or program parameters in a consistent way. COM-PACK provides menu trees with well-defined formatsand user interactions . Both line mode and full-screenmodes are available .

Packages are provided within COMPACK for thecreation of new menus and modifications of existingmenus. Various utilities are incorporated allowingtransparent handling of data, display and menu specifi-cation files, parameter specification, and fielding ofsystem messages .

7.2.4. Graphics and displaysThe DO event display facility uses the graphics

system DI3000 [126], linked with a package written byDO called PIXIE. The package gives the capability tomake two- or three-dimensional views of detectors withsuperimposed hit or reconstructed track/cluster infor-mation, or Lego plots of calorimetric energy deposits .A special package has been made to display the pri-mary Monte Carlo events from ISAJET or other gener-ators . The routines of this package are available forinclusion into any of the on-line or off-line DO pro-gram frameworks . Special facilities are provided inPIXIE for temporary or permanent zoom, hardcopy, orreal-time editing of screen parameters .

Histogramming is done in DO using the CERNpackage HBOOK4. Display of histograms is done withHPLOT5 (the CERN display package), PAW (CERNphysics analysis workstation package) or DISPLAYS (aDO modification of an LBL display package) . DO usesthe Fermilab-maintained DI-3000 version of thePAW-HIGZ package.

Three-dimensional displays of events are also madeon Evans and Sutherland [127] color workstations per-mitting real time rotation, pan, and zoom operations .

7.2.5. Program builderA general facility has been developed for assem-

bling DO programs which simplifies and standardizes

many routine functions . For example, reading events,creating standard output files, exercising certain stan-dard analysis packages in the correct order, etc . i sbetter kept hidden from the ordinary user so that theseprocesses remain uniform and controlled . The tool forassembling the standard building blocks and enforcingdiscipline among users is the program builder (PBD).A standard framework is provided for each generalfunctional group of programs (e .g . Monte Carlo simu-lators, event reconstruction, on-line event monitoring,and offline event analysis).

The program builder is driven by easily editableASCII files . For each framework, there is a file listingthe "user hooks", and files for every application pack-age listing the corresponding interfaces to be called bythose hooks. The hooks are typical stages in programs,such as job initialization, event processing, event dis-play, event dump, job summary, etc. In addition to thelist of interfaces, such package files also list any specialapplication object libraries and any RCP files theyrequire.A user can easily create a new package file by

editing a previously existing package file . The sameapplication package can be used in different frame-works. Thus, a package developed in the off-line recon-struction may also be used in an on-line framework.The program builder is invoked by a simple commandgiving the framework and list of application packagesas options. It then generates the Fortran code for thecalls to interfaces by the user hooks, a link commandfile to make the executable, and a setup command fileto run it . Complex application packages can be builtfrom combinations of simpler application packages .Communication between applications can only occurthrough ZEBRA banks or utilities that access ZEBRAbanks.

7.3. Event Monte Carlo

The Monte Carlo simulation of the DO experimenthas been of great importance throughout the planningand building phase of the project, and of course isessential for performing physics analyses . Early mod-elling gave crucial information affecting the detaileddesign of the detector . For example, the design of theinterface between central and end calorimeters wasone of the most difficult to understand and the MonteCarlo simulations here were invaluable . In addition,preparation of the full reconstruction code for theanalyses of events required a simulation of the fulldetector with all detailed physical processes of scatter-ing and interactions included . This simulation, carriedto the level of raw ADC hit simulation, was used todevelop the event reconstruction code and to test itsperformance.

The DO Monte Carlo [128] is based on the GÉANT

DO Collaboration INucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

package developed at CERN [129] . This package pro-vides tools for specifying volumes containing particularmaterials and the framework for transporting particlesthrough these volumes with appropriate physical scat-tering and interaction processes included . Processessuch as A-ray production, multiple Coulomb scattering,full showering by electromagnetic and hadronic parti-cles, decays, and Bremsstrahlung by electrons andmuons are accurately simulated in GEANT. The en-ergy deposits in any specific volume are recorded forsubsequent collection and transformation to digitizedsignals .

'~~ l' 'il~ll'lûll

The geometric simulation for the DO detector isquite detailed . The tracking chambers and muon cham-bers are simulated in full detail down to the level ofthe sense wires, cathode material, support structures,etc . Fig . 83 shows the tracks from a representative fitevent in which the various physical processes give riseto non-ideal straight-line tracks . The calorimeter mod-elling is less complete owing to the complexities of theshowering process and the consequent CPU require-ments : the full structure of supports and individualmodules are present but the internal absorberplate/liquid argon structure is replaced with a mixture

Fig . 83 . r-z view of the tracks in the DO tracking detectors and beginning of the central calorimeter from a single t-i event .

24 5

24 6 DO Collaboration INucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

of suitable effective atomic weight . Sampling fluctua-tions are added after showering for each track, and theappropriate hadron to electron response (as deter-mined from DO tests discussed in Section 3.7) is intro-duced. The appropriate Ovl X AO X Az segmentationof the readout cells is imposed and the compensatedenergies are added for each cell . For most applications,the electromagnetic showers are allowed to evolve untilthe particle energies fall to 200 MeV; beyond this pointenergies are estimated from simple parametrizations[130] resulting in a time saving of approximately afactor of 3. The energy deposits in the calorimeter cellsare summed over all particles and noise is added.

For test beam studies, the simulation techniqueusing calorimeter material mixtures has been com-pared with calculations using a full plate simulation, aswell as with single electrons and pions data . Fig. 84shows a typical hadronic shower in the IH module ofthe end calorimeter as simulated in the plate level testbeam Monte Carlo. Fig. 85 shows the comparison ofelectron and pion energy resolutions obtained for bothversions of Monte Carlo and data ; both Monte Carlotechniques give adequate representation for the data .Fig. 86 shows the distribution of the fraction of energydeposited in the EM4 layer of the CC for 50 GeVelectrons near 90°. The results from test beam data,

Fig . 84 . Simulated hadronic shower for a 100 GeV pion in the end IH calorimeter module.

Yb

N

T

CWOCO

ali

Do Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

247

0.04

0.02

0 20 40 60 80 100 120 140 160 180Beam Energy (GeV)

and the Monte Carlo using the full plate representa-tion and the mixture Monte Carlo with full showeringand shower parametrization, are shown. The plateMonte Carlo is a good representation to the data; theparametrized mixture Monte Carlo is less good, butadequate for most purposes . Fig. 87 shows the compar-ison of energy deposited in the ECIH hadron calorime-ter layers for 100 GeV pions at 71 = 1.95. Both testbeam data and the plate Monte Carlo distributions areshown and the agreement is good . Though not a per-fect representation of all aspects of the data, the mix-ture version of the Monte Carlo for the calorimeters is

120 140 160Beam Energy (GeV)

Fig. 85 . (a) Fractional energy resolution vs energy for electrons in the ECEM for test beam data, mixture Monte Carlo and plateMonte Carlo. The curve is a fit to the data. (b) Fractional energy resolution vs energy for pions in the EC for test beam data, withand without a veto on ECEM Layer 1 energy of E <0.51 GeV, and for the plate Monte Carlo. The curves are fits to the data

distributions .

useful because it allows a reduction in CPU time forsingle particle event generation by a factor of over fourwith respect to the full representation of the calorime-ter at the plate level. For simulation of i-t events inthe full DO detector, the mixture representation MonteCarlo requires about 23 VUP hours per event.

Another technique for speeding up the full MonteCarlo simulation of p-p collisions which has been usedsuccessfully involves the creation of libraries [131-133]of showers in the calorimeters, generated from the fullGEÀNT simulation . These showers are parametrizedin terms of the particle momentum (seven ranges),

248

location of the primary vertex along the beam line (sixranges) and o7 (37 ranges) for both EM and hadronicshowers. Interpolation of variables within the range isperformed when a shower with approximately correctvariables is chosen . The number of showers in thelibrary is approximately 1.3 X 10 6 . The shower librarysimulations have been extensively compared with thefull (mixture level) event Monte Carlo simulations andrepresent the distributions and correlations quite well[133] . Simulation of events with shower libraries re-quires about 25 VUP-s for t-t events .

7.4. Off-line event reconstruction and processing

Development of the full DO reconstruction pro-gram DSÖRECO has proceeded through the steps out-lined above. Extensive SA/SD specifications weredrawn up and discussed . ZEBRA structures were de-signed for data from each of the detectors, with in-creasing complexity of calculation and refinement . Thecode has been continuously developed by a large num-ber of physicists working in many locations, with fre-quent updates and releases of the several libraries .

400

300

200

100

400

300

200

100

Do Collaboration /Nuel. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

50 GeV Electrons (77=0.05),

Fractional Energy EM Layer 4

Data

D 1

I

IF"_L-, , I .~ I

~I0 0.2 0.4 0.6 0.8 1

EM4/Etotal

Monte CarloMixturesFull Showering

0 0.2 0.4 0.6 0.8 1EM4/Etotal

Certain rules were adopted from the beginning. Allphysics and calibration data were stored in ZEBRAbanks. The various sub-packages of DORECO commu-nicate with each other only through ZEBRA blocks,and a strictly limited number of interface routines.Although most code development and the DO librariesare accomplished on VAX-VMS machines, all codehas been developed in FORTRAN 77, augmented by alimited number of specific extensions . Special ruleswere adopted for machine dependent code . A toolcalled DOFLAVOR was developed which checked allcode for adherence to DO standards. A special editor,EVEDT [134], was employed which supported stan-dard conventions for programming. EVEDT also im-plemented methods for accessing the source for anysubroutine and ZEBRA bank documentation in theDO library by simple request key operations . A toolwas developed for accessing ZEBRA bank documenta-tion and content in memory during a debug session.The detector geometry and calibration constants arekept in RCP files . In applications programs, the FZ orRZ files made from these RCP files are accessed andmaintained via the database program DBL3 .

400

300

200

100

400

300

200

100

Monte CarloPlates

.I I

0.2 0.4 0.6 0.8 1EM4/Etotal

Monte CarloMixturesParametrization

I

a

, I

. i'r~._rl

, I

' ~. I ~

l0 0.20.4 0.60.8 1

EM4/Etotal

Fig. 86 . Energy distributions for 50 GeV electrons near 90° in the CCEM layer 4 for (a) test beam data ; (b) the plate Monte Carlo ;(c) the mixture Monte Carlo with full showering; and (d) the mixture Monte Carlo with shower parametrization below 200 MeV.

The detector specific code resulting in CD tracks,energy deposits in calorimeters, vector energy sums,and muon tracks was developed mostly by physicistsclosely connected with each detector . Once this codewas completed in its first version, new groups wereformed for the preparation of algorithms for physicssignatures - electrons, photons, muons, jets and miss-ing ET . Membership in these groups was generallyrather different from the original detector code devel-opment . The full DORECO package contains about150 000 lines of code, exclusive of utility libraries .

All DO physics data is logged to double density 8mm video-quality magnetic tapes . The data samplesexpected for an approximately 12 month run are ex-pected to be of order 20 million events and 10 Ter-abytes . In addition, those events judged to be of high-est interest (approximately 10% of the full sample)based on Level 2 filter information are sent to the"express line" over Ethernet for immediate reconstruc-tion by one of six VAX 3100/76 workstations clusteredwith the host computer . Event analysis of the full dataset with DORECO is made using the Fermilab CentralComputing Facility farm of silicon graphics 4D/35processors . A silicon graphics 4D/430 server provides

DO Collaboration /Nuel. Instr. and Meth. in Phys . Res. A 338 (1994) 185-253

100 GeV Pions at i7=1 .95

.'!TLw~-

'

_ J I

40 80 120

E (GeV)

program and calibration database storage . The produc-tion capability of this processor farm is about 100000events per day and is sufficient to keep processing upto date .Two output data files from each run with DORECO

are provided : the STA file contains the raw data plusthe complete results of reconstruction . The STA eventsare used for reprocessing events of particular interestand in preparing displays . Each STA event containsapproximately 600 kbytes . The DST file contains acompressed version of the event with the most impor-tant information intended to serve the bulk of subse-quent analysis needs . The DST contains summaries oftrack and clusters as well as all parameters for elec-trons, photons, muons, jets, taus and missing ET. TheDST event size is approximately 20 kbytes . Both ver-sions utilize identical ZEBRA structures .

Control of the event reconstruction is providedthrough a software tool called the production manager.These programs supervise and control the flow ofproduction jobs on each of the operations platforms .Multiple projects on multiple nodes are managed andprovided with the necessary resources . Tools have beenprovided to submit and monitor jobs and to accumu-

d

'~°0-�. l .

I

I0 40 ' I 80 ' 120

E (GeV)

249

Fig. 87 . Energy deposition in the ECIH calorimeter layers for 100 GeV pions at 71 =1 .95 for test beam data and the plate MonteCarlo.

250 200IH 1 1112- Data - Data

200 MC Plates . . MC Plates. . . 150

150

100100

5050

.40 80 120 40 80 1 120

E (GeV) E (GeV)

400IH3 IH4- Data 800 - Data. . . MC Plates . . . MC Plates

300600

200400

100 200

250 DO Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

late statistics on jobs . Each production job is fullyspecified in dedicated RCP files ; submission of a job isdone in consultation with a global resources RCPtable.

The history of all files produced by the host com-puter and the off-line production system is kept in aproduction database (PDB), based on RdB [114] . Inaddition to all types of event files, PDB keeps informa-tion on tapes, production programs, trigger tables andthe relevant relationships . Dedicated servers trap in-formation from the host computer and the productionmanagers and insert it into the PDB. Tools have beendeveloped to enable queries and make lists of datasetsconforming to specified conditions .

All files produced by the production manager areassigned generic names uniquely identifying the dataset,as well the production path . All DO frameworks canaccept these generic names for input files . Physicsapplication programs are provided with requesteddatasets through the CERN distributed file and tapemanagement facility FATMEN [135] by referencingthese generic names. These references are typicallymade across the boundaries between analysis and fileserving/storage clusters .

The STA files are typically kept on tape, while theDSTs are largely kept on disk of a special file servingcluster consisting of VAX 4000/60 and 3100/76 work-stations . This cluster provides approximately 300Gbytes of disk storage and sixteen 8 mm tape drives .Robotic tape mounts provide for 128 tapes on two ofthe tape server nodes. This cluster, housed in the maincomputing facility, is split into nine Ethernet segmentsand is connected through a FDDI network to theremote physics analysis clusters .

In addition to serving files for analysis, the fileserver cluster is used to run I/O intensive streamingand filtering jobs . Four overlapping data streams aredefined on the basis of the Level 2 filter bits ; they areelectrons, muons, mininum bias, and missing ET. Thesestreams are provided for both STA and DST files andare intended to streamline any reprocessing made nec-essary by reconstruction program changes. In addition,events are filtered following reconstruction, based uponthe full available off-line information . The pool offiltered events consists of approximately 25% of theoriginal sample for all physics groups except QCDjets .The QCD events are handled separately since most ofthese triggered events are of good quality. Instead offiltering these, special "micro DST" files are writtenwhich achieve a factor of three more compression thanthe regular DSTs.

The off-line reconstruction DQRECO was availablewell before the first DO data from the collider . As ameans of checking the code and allowing its evolution,a set of 100 000 events were prepared with the DOMonte Carlo DQGEANT. These events were repre-sentative of what would be obtained with 33 nb -1 of

data collection. They mainly contained the dominantQCD hard parton scattering processes with thresholdsfor jet triggering set to 50 GeV/c. In addition, someselected physics events of rarer provenance were added(together with triggers designed to select events withleptons and/or missing ET). Analysis of this sampleand associated known background samples served asthe vehicle for developing DORECO and its algo-rithms, as well as helping to guide the evolution of thevarious physics analysis techniques .

Acknowledgements

We express our gratitude to the superb supportstaffs at each of the DO collaborating institutions . Thetechnical staff within the Accelerator, Research andComputing Divisions and the Physics and TechnicalSupport Sections at Fermilab have played vital roles indesign, fabrication, test and operation of the detector .Financial support for the DO detector has been pro-vided by the U.S. Department of Energy, the USSRState Committee for Atomic Energy, the Commissariatà L'Energie Atomique in France, the U.S . NationalScience Foundation, and the Department of AtomicEnergy in India.

References

[1] Design Report, The DO Experiment at the Fermilabp-p Collider, November 1984.

[2] S.H . Aronson et al., Proc . Conf . on Multiparticle Dy-namics, Kiryat Anavim (1985) p. 827.

[3] P. Franzini, Proc . XXth Rencontre de Moriond, LesArcs, France, (1985) p. 185 .

[4] R. Yamada, Proc . on Physics of Proton AntiprotonCollisions, Tsukuba, Japan (1985) p. 371.

[5] M.A . Abolins, Proc . First Aspen Winter Physics Conf.,Ann. N.Y . Acad . Sci . 46 1 (1986) 63 .

[6] P.D . Grannis, Proc . Les Rencontres de Physique de laVallée d'Aoste, La Thuile, Italy (1987) p. 253.

[7] A.P . White, Proc . 8th Topical Workshop on Proton-Antiproton Collider Physics, Castiglione, Italy (1989).

[8] R. Madaras, Proc . Division of Particles and Fields Con-ference, Batavia, IL (November 1992) .

[9] A. Clark et al ., Nucl . Instr . and Meth . A279 (1989) 243.[10] D. Buchholz et al., Nucl. Instr. and Meth . A 257 (1987)

556.[111 A.R . Clark et al ., Nucl . Instr . and Meth . A 261 (1987)

420.[12] A.R . Clark et al ., Nuel . Instr . and Meth. A 315 (1992)

193.[13] NiCoTin is an alloy of Co, Cr, Ni, Fe, Mo in respective

fractions of 38 .8%, 19.3%, 15 .6%, 13 .3%, and 11 .4%,manufactured by Microfil Industries SA, Switzerland .

[14] J. Kadyk, Nucl . Instr . and Meth. A 300 (1991) 436.[15] V.L . Ginzberg and I.M . Frank, JETP 16 (1946) 15 .

DO Collaboration /Nucl. Instr. and Meth . in Phys. Res. A 338 (1994) 185-253

[16] J.-F. Detoeuf et al ., Nucl . Instr . and Meth . A 265 (1988)157 .

[17] J.-F. Detoeuf et al ., Nucl . Instr . and Meth . A 279 (1989)310 .

[18] F . Feinstein, Ph.D Thesis, A l'Universite de Paris-Sud,Centre d'Orsay, December 1987 (unpublished) .

[19] T. Behnke, Ph.D Thesis, State University of New Yorkat Stony Brook, August 1989 (unpublished) .

[20] R . Avery et al ., IEEE Trans. Nucl . Sci . NS-40 (1993)573 .

[21] R . Yarema, IEEE Trans . Nucl . Sci . NS-33 (1986) 933 .[22] B . Chase, DO Note #953 (Apr . 1990) (unpublished).[231 G . Saewert and B . Chase, DO Note #669 (Jan. 1988)

(unpublished).[24] D . Buchholz, D . Claes, B . Gobbi and S . Park, DO Note

#326 (Dec . 1985) (unpublished) .[25] M.I . Martin, M.E . Johnson, M.J . Mayberry and D.C.

DeGroot, IEEE Trans . Nucl . Sci . NS-34 (1987) 258 .[26] D . DeGroot, DO Note #1013 (Aug . 1990) (unpub-

lished).[27] S . Rajagopalan, Ph.D . Thesis, Northwestern University,

June 1992 (unpublished) .[28] J . Bantly, Ph.D . Thesis, Northwestern University, June

1992 (unpublished) .[29] Y . Ducros et al ., Nucl . Instr . and Meth . A 277 (1989)

401 .[30] W . Guryn, Proc. Second Int. Conf. on Calorimetry in

High Energy Physics, Capri, Italy, Oct 14-18, 1991 .[311 Manufacturing Sciences Corp., Oak Ridge, TN .[321 M.M . Baker et al ., Trans. Faraday Soc . 62 (1966) 2525 .[33] NVF Company, P.O . Box 6096, Broadview, IL 60153.[34] Abatron Inc ., 33 Center Dr., Gilberts, IL 60136 .[35] R.D . Luther, G.T. Mulholland and K.J. Krempetz, Adv .

Cryogen . Eng . 35B (1989) 1819 .[36] A.L. Spadafora, Proc . 5th Pisa Meeting on Advanced

Detectors, La Biodola, Isola d'Elba, Italy, May 26-31,1991, eds . A. Baldini, A . Scribano and G . Tonelli, Nucl .Instr . and Meth . A 315 (1992) 279 .

[37] H . Aihara et al., Nucl . Instr . and Meth . A 325 (1993)393 .

[38] Bicron Corp ., 12345 Kinsman Rd ., Newbury, OH 44065 .[39] The small-I 771 boxes, where the corrections are expected

to be smaller, employed Russian built 10-stage PM60phototubes . In the large~,11 boxes where dead materialcorrections are expected be largest, we have usedHamamatsu R647 phototubes .

[40] G.C. Blazey, Proc . First Int . Conf. on Calorimetry inHigh Energy Physics, Batavia, IL, Oct 29-Nov 1, 1990,eds. D . Anderson, M. Derrick, H.E. Fisk, A . Para andC. Sazama, p . 101 .

[41] W . Hoffmann et al ., Nucl. Instr . and Meth . 135 (1976)151 .

[42] E .I. DuPont de Nemours Inc ., Wilmington, DE .[43] P. Franzini, Proc. First Int. Conf. on Calorimetry in

High Energy Physics, Batavia, IL, Oct 29-Nov 1, 1990,eds . D . Anderson, M . Derrick, H.E . Fisk, A. Para andC . Sazama, p. 359.

[44] R.D . Schamberger, Proc . Second Int. Conf. on Ad-vanced Technology and Particle Physics, Como, Italy,11-15 June, 1990 .

251

[45] M . Demarteau, Proc. First Int . Conf. on Calorimetry inHigh Energy Physics, Batavia, IL, Oct 29-Nov 1, 1990,eds . D. Anderson, M. Derrick, H.E . Fisk, A . Para andC . Sazama, p . 91 .

[46] P.M . Tuts, Proc . on Instrumentation For Colliding BeamPhysics, Novosibirsk (1985), p . 183 .

[47] S. Feher, Proc . Second Int . Conf. on Calorimetry inHigh Energy Physics, Capri, Italy, Oct 14-18, 1991 .

[48] D.L. Huffman, Proc. IEEE Meeting, Nov . 1991 (SantaFe, NM) .

[49] B . Cox et al., Proc . 1985 DPF Meeting of the AmericanPhysical Society, Eugene, Oregon, p . 887 .

[50] P.M. Tuts, Proc. 1986 Conf. on High Energy Physics,Berkeley, CA, 1455 (1986) p . 1455 .

[51] S . Aronson et al ., Nucl . Instr . and Meth . A 269 (1988)492 .

[52] Paolo Franzini, Nucl . Instr. and Meth . A 263 (1988) 78 .[53] S .J . Wimpenny et al ., Nucl . Instr . and Meth . A 279

(1989) 107.[54] M . Abolins et al ., Nucl . Instr . and Meth . A 280 (1989)

36 .[55] P . Draper, Proc . European Physical Society, Madrid,

September 1989, and Nucl . Instr . and Meth . B 16 (1990)506.

[561 P . Franzini, Nucl . Instr. and Meth . A 289 (1990) 438.[57] S . Abachi et al., Nucl . Instr . and Meth . (1993) 53.[58] A.L . Spadafora, Proc . First Int . Conf. on Calorimetry in

High Energy Physics, Batavia, IL, Oct 29-Nov 1, 1990,eds . D . Anderson, M. Derrick, H.E. Fisk, A . Para andC . Sazama, p . 77 .

[59] H . Aihara, Proc . 1990 IEEE Nuclear Science Sympo-sium (Arlington, VA) Oct . 23-26, 1990.

[60] N.A . Amos, Proc . First Int . Conf. on Calorimetry inHigh Energy Physics, Batavia, IL, Oct 29-Nov . 1, 1990,eds. D . Anderson, M . Derrick, H.E . Fisk, A. Para andC. Sazama, p . 84 .

[61] N.A. Roe, Proc. Second Int . Conf. on Calorimetry inHigh Energy Physics, Capri, Italy, Oct . 14-18, 1991 .

[62] J. Kotcher, Ph.D . thesis submitted to New York Univer-sity, October 1992 (unpublished).

[63] These studies have recently been extended down to 2GeV .

[64] R . Wigmans, Nucl. Instr . and Meth . A 259 (1987) 389 ;ibid . A 265 (1986) 53 .

[65] R . Engelmann et al ., Nucl . Instr. and Meth . 216 (1983)45 .

[66] D . Kewley, G.C. Blazey, J . Chiu and T . Ferbel, Nucl .Instr . and Meth. A 290 (1990) 346 .

[67] K . De, DO Note #1017 (Aug . 1990) (unpublished).[68] R . Stephens, DO Note #1165 (Sept . 1991) (unpub-

lished).[691 D . Green et al., Nucl . Instr . and Meth . A 244 (1986)

356 .[70] R . Yamada et al ., IEEE Trans . Magn. MAG-28 (1992)

520 .[71] R . Yamada et al ., Fermilab TM-1786 (1992) .[72] C. Brown et al ., Nucl . Instr . and Meth . A 279 (1989)

331.[73] Glasteel Inc ., 1727 Buena Vista Ave, Duarte, CA .

252 Do Collaboration /Nucl. Instr. and Meth. in Phys. Res. A 338 (1994) 185-253

[74] D . Green et al ., Fermilab TM-1523, May 1988 (unpub-lished) .

[75] J.M. Butler et al ., Nucl. Instr . and Meth . A 290 (1990)122 .

[76] Yu.M . Antipov et al ., Nucl. Instr . and Meth . A 297(1990) 121 .

[77] Mao Huishun, DO Note #1014 (Aug . 1990) (unpub-lished).

[78] T . Marshall, DO Note #761 (Sept . 1989) (unpublished) .[791 N . Oshima and S . Igarashi, DO Note #805 (Feb . 1989)

(unpublished) .[80] G.S . Gao and R . Partridge, IEEE Trans. Nucl . Sci .

NS-38 (1991) 286 .[81] M. Abolins et al ., IEEE Trans . Nucl . Sci . NS-36 (1989)

384 .[82] M. Abolins et al ., Nucl . Instr. and Meth . A 289 (1990)

543 .[83] M. Fortner et al., IEEE Trans . Nucl . Sci. NS-38 (1991)

480 .[841 M. Utes, M . Johnson and M . Martin, Proc . IEEE Meet-

ing, Nov. 1991 (Santa Fe, NM).[85] D . Cutts, R. Zeller, R.D . Schamberger and R . Van

Berg, Proc . 1984 Conf. on Recent Developments InComputing, Processor and Software Research ForHigh-energy Physics, Guanajuato, Mexico, p . 403 .

[86] D . Cutts, et al., Proc . Conf. on Computing In HighEnergy Physics, Amsterdam (1985) p . 287 .

[87] D . Cutts et al ., Proc . IEEE 1988 Nuclear Science Sym-posium, Orlando, FL, Nov. 9-11, 1988 .

[88] D . Cutts et al., Proc. Int . Conf . on Computing in HighEnergy Physics, Oxford, England, Apr. 10-14, 1989 .

[891 D. Cullen-Vidal et al ., Proc . Conf. on Computing inHigh Energy Physics, Tsukuba, Japan (March 1991).

[901 D. Cutts et al ., Proc. Conf. on Computing in HighEnergy Physics, Annecy, France, CERN Preprint 92-01,p . 262 (1992) .

[91] ZRL Inc ., Bristol, RI .[921 R. Angstadt et al ., Proc IEEE Meeting, Nov . 1991

(Santa Fe, NM) .[93] J . Hoftun, DO Note #909 (Dec . 1989) (unpublished) .[94] J .S. Hoftun, et al ., IEEE Trans . Nucl . Sci. NS-34 (1987)

710.[95] D . Cutts and J .S . Hoftun, Conf . on Computer Simula-

tion Techniques in High Energy Physics, Argonne, Ill .,Aug . 1987 .

[96] H . Stredde, DO Note #740 (Aug . 1988) (unpublished).[97] S .J . Wimpenny et al ., Nucl . Instr . and Meth . A 279

(1989) 359 .[981 M.-J . Yang et al ., Proc . Second Int . Conf. on Advanced

Technology and Particle Physics, Como,Italy, July 11-17,1990, University of California at Riverside preprintUCR/DO/90-10 (1990) .

[99] S .-C . Ahn et al., Proc . IEEE Meeting, Nov . 1991 (SantaFe, NM).

[100] C . Rotolo, DO Note #926 (Jan . 1990) (unpublished);C . Rotolo and M . Fachin, DO Note 927 (Jan . 1990)(unpublished) ;S .J . Chappa and C . Rotolo, DO Note #1214 (Oct. 1991)(unpublished) .

[101] L . Paterno and J .F . Bartlett, DO Note #1029 (Oct .1991) (unpublished).

[102] H . Jöstlein, Proc . Second Int . Workshop on the Align-ment of Accelerators, Hamburg, Germany, September1990 .

[103] Capacitec, P.O . Box 819, Ayer, MA.[104] The Fredericks Co ., P.O . Box 67, Huntington Valley,

PA .[105] Honeywell Corp. model 924AB4W-L2P .[106] R . Goodwin et al ., Nucl . Instr. and Meth . A 247 (1986)

107 .[107] R. Goodwin et al., Proc . Int. Conf. on Accelerator and

Large Experimental Physics Control Systems, Vancou-ver, British Columbia, Canada, October 1989, p. 125 .

[108] M.F . Shea et al ., Proc . Europhys . Conf . on ControlSystems for Experimental Physics, Villars, Switzerland,September 1987 .

[109] J .F . Bartlett, S. Fuess and H . Prosper, DO Note #866(Dec . 1989) (unpublished).

[110] M . Frey and A. Waller, DO Note #821 (Dec. 1988)(unpublished) .

[111] M . Frey, DO Note #822 (Dec. 1988) (unpublished) .[112] M . Frey, DO Note #823 (Dec . 1988) (unpublished) .[113] M. Frey and A. Waller, Note, DO Note #824 (Dec.

1988) (unpublished) .[114] Digital Equipment Corp ., Maynard, MA 01754 .[115] S . Fuess, DO Note #1558 (Nov . 1992) (unpublished).[116] P . Bhat, Beam, Load 2, DO Note #1104 (Mar . 1991)

(unpublished) .[117] C . Grozis et al., Adv. Cryogen . Eng . 35 (1990) 1811 .[118] M . Page-Jones, The Practical Guide to Structured Sys-

tems Design (Yourdon Press, 1980).[119] S . Mellor and P . Ward, Structured Development for

Real Time Systems (Yourdon Press, 1986).[120] G . Kellner, private communication and ALEPH notes

127 and 142 .[121] J . Linnemann et al., Comput . Phys . Commun . 45 (1987)

245 .[122] M. Goosens et al., CERN Library Program No . 0100

(Nov. 1991) .[123] B. Adeva et al ., Nucl . Instr . and Meth . A 309 (1991) 318 .[124] H . Prosper, Proc . First Int . Conf. on Calorimetry in

High Energy Physics, Batavia, IL, Oct 29-Nov 1, 1990,eds . D . Anderson, M . Derrick, H.E . Fisk, A . Para andC . Sazama, p . 546 .

[125] J . Hoftun, DO Notes #542 and #543 (Apr . 1987) (un-published).

[126] DI3000 Users Guide, Visual Numerics, Inc., 14141Southwest Freeway, Suite 3000, Sugarland, TX 77478 .

[127] Model PS390 3D display terminal, Evans and Suther-land, 540 Arapeen Drive, PO Box 58700, Salt Lake City,UT 84158 .

[128] A.M . Jonckheere, Proc . Argonne National LaboratoryDetector Simulation Workshop, August 1987 and DONote 638.

[129] F . Carminati et al ., GEANT User's Guide, CERN Pro-gram Library (Dec . 1991) .

[130] S . Linn, DO Note #874 (Aug. 1989) (unpublished).

DO Collaboration /Nucl. Instr. and Meth. in Phys . Res . A 338 (1994) 185-253

[131] W.J. Womersley, R. Raja and A.M . Jonckheere, DONote #650 (Dec. 1987) (unpublished) .

[132] R . Raja, Proc. Workshop on Detector Simulation forthe SSC, Argonne, IL, Aug . 24-28, 1987 .

[133] N.A. Graf, Proc. First Int. Conf. on Calorimetry in HighEnergy Physics, Batavia, IL, Oct . 29-Nov. 1, 1990, eds .

253

D . Anderson, M . Derrick, H.E . Fisk, A . Para and C .Sazama, p . 539.

[1341 O . Callot, DO Notes #564 (June 1987) and #728 (July1988) (unpublished).

[135] J .D . Shiers, FATMEN, CERN Program Library, 1992.