EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH

CERNPPE

February

The CPLEAR Detector at CERN

The CPLEAR Collab oration

R Adler WAlbrecht TAlhalel A Angelop oulos A Ap ostolakis E Aslanides

G Backenstoss F Bal JP Bard D Barraca CP Bee O Behnke

A Benelli J Bennet V Bertin F Blanc P Blo ch MBonnet C Bula

A Calzas P Carlson M Carroll JCarvalho ECawley S Charalamb ous

M Chardalas G Chardin P Charra MB Chertok ACody J da Silva

D Damianoglou R Daniel M Danielsson P Dechelette M Dedieu

S Dedoussis M Dejardin J Derre A Dijksman BDinkespiller MDodgson

M Droge J Duclos J Dudragne D Durand A Ealet B Eckart

C Eleftheriadis CEngster IEvangelou LFaravel PFassnacht JLFaure

C Felder R FerreiraMarques W Fetscher M Fidecaro A Filipcic

D Francis J Fry C Fuglesang E Gabathuler Y Gally R Gamet

D Garreta D Geiss T Geralis HJ Gerb er AGo P Gumplinger

D Guyon C Guyot P Harrison PF Harrison A Haselden PJ Hayman

E Hazen F HenryCouannier WG Heyes RW Hollander E Hub ert

C Jacobs K Jansson HU Johner K JonAnd NKarkour A Kerek

G Kesseler PR Kettle DKing T Klados CKochowski PKokkas

K Kontek R Kreuger TLawry T Lecouturier RLeGac F Leimgrub er

D Linget A Liolios B Lofstedt F Louis EMachado PMaley U Mall

I Mandic NManthos GMarel CP Marin H Martin JC Michau

agels M Mikuz J Miller FMontanet TNakada F Nanni A Onofre BP

I Papadop oulos PPavlop oulos FPelucchi PPetit K Philipp oussis J Pinto

da Cunha APolicarp o GPolivka HPostma Ch Rheme RRickenbach

BL Rob erts E Rozaki T Ruf LSacks LSakeliou P Sanders C Santoni

K Sarigiannis MSchafer LA Schaller TSchietinger ASchopp er PSchune

A Soares M Steinacher STatsis LTauscher C Thibault FTouchard

C Touramanis F Triantis DA Troster I Tsamouranis H Tschopp

P Tsilimigras EVan Beveren CWE Van Eijk BVan Koningsfeld JPVanuxem

G Varner HVerweij S Vlachos DWarner EWatson PWeb er HWendler

O Wigger C Witzig MWolter CYeche DZavrtanik and D Zimmerman

Submitted to Nucl Instrum Methods A

UniversityofAthens GR Athens Greece

University of Basle CH Basle Switzerland

Boston University Boston MA USA

CERN CH Geneva Switzerland

LIP and University of Coimbra P Coimbra Portugal

Delft UniversityofTechnology JB Delft Netherlands

UniversityofFrib ourg CH Frib ourg Switzerland

University of Ioannina GR Ioannina Greece

UniversityofLiverp o ol Liverp o ol L BX UK

J Stefan Inst and Phys Dep University of Ljubljana SI Ljubljana Slovenia

CPPM INPCNRS et Universite dAixMarseille I I F Marseille France

CSNSM INPCNRS F OrsayFrance

PaulScherrerInstitutPSI CH Villigen Switzerland

CEA DSMDAPNIA CESaclay F GifsurYvette France

Royal Institute of Technology KTH S Sto ckholm Sweden

University of Thessaloniki GR Thessaloniki Greece

ETHIPP Zuric h CH Zuric h Switzerland

Abstract

The CPLEAR collab oration has constructed a detector at CERN for an extensive programme of

CP T and CPTsymmetry studies using K and K pro duced by the annihilation ofps in a

hydrogen gas target The K and K are identied by their companion pro ducts of the annihila



tion K which are tracked with multiwire prop ortional chamb ers drift chamb ers and streamer

tub es Particle identication is carried out with a liquid Cherenkov detector for fast separation of

pions and kaons and with scintillators whichallow the measurement oftimeofight and energy

loss Photons are measured with a leadgas sampling electromagnetic calorimeter The required

antiproton annihilation mo des are selected by fast online pro cessors using the tracking chamber

and particle identication information All the detectors are mounted in a T uniform eld of

an axial solenoid of diameter m and length m to form a magnetic sp ectrometer capable of full

online reconstruction and selection of events The design op erating parameters and p erformance

of the subdetectors are describ ed

Intro duction

The aim of the CPLEAR exp eriment is to study CP T and CPTsymmetries in the

neutral kaon system The exp erimental metho d consists of measuring asymmetries b etween

the decay rates of K and K into various nal states f f

as a function of the decayeigentime

RK f RK f

A

f

RK f RK f

The exp eriment installed at the Low Energy Antiproton Ring LEAR at CERN uses initially

pure K and K states pro duced concurrently in the annihilation channels

pp K K

pp K K

each with a branching ratio of The strangeness of the neutral kaonistaggedbythe

charge of the accompanying kaon and is therefore known eventbyevent The momentum of the

 

pro duced K K can b e obtained from measurement of the K pair kinematics

The detector sp ecications are based on the exp erimental requirements which are

To select reaction from the very large number of multipion annihilation channels

In particular very ecientkaon identication is essential

To distinguish b et ween the various neutral kaon decaychannels

a

To measure the decay eigentime b etween and K mean lives At the highest K

S

momentum measured in our exp eriment MeVc the K mean decay length is cm

S

This sets the size of the cylindrical K decayvolume to a radius of cm

To acquire a large quantity of statistics which requires b oth a high rate capabilityMHz

annihilation rate and large geometrical coverage

An imp ortant asp ect in the design of the exp eriment is the need to minimize neutral kaon

regeneration eects in the decayvolume by minimizing the amount of matter in the detector The

regeneration eects mo dify the time evolution of initial K and K dierently The regeneration

amplitude has not b een measured in our K momentum range and must b e inferred from previous

measurements of charged kaon crosssections

Since the antiproton reaction is observed at rest the particles are pro duced isotrop

ically and the detector has a typical near geometry The whole detector is emb edded in a

m long m diameter warm solenoidal magnet whichprovides a T uniform eld This

magnet was previously used in the DM e e exp eriment at Orsay

wn in Fig The antiprotons are The general layout of the CPLEAR exp eriment is sho

stopp ed in a pressurized hydrogen gas target The use of liquid hydrogen was ruled out in order

to minimize the amount of matter in the decayvolume The high pressure of the hydrogen

bar and the low momentum of the incomingpbeamhelptok eep the size of the stopping region

small

a

Unless explicitly stated wemeanbyK either K or K t out of the CPLEAR exp erimen y

CPLEAR Detector Figure The general la

Magnet coils A support rings

Electromagnetic calorimeter Drift chambers 1 m Beam monitor 16 bar H2 target 200 MeV/c p Proportional chambers Streamer tubes

Cherenkov and scintillator counters

A series of cylindrical tracking detectors provides information ab out the tra jectories of

charged particles charged tracks or tracks in order to determine their charge signs momenta

and p ositions A mo derate momentum resolution pp between and is sucientto

p erform the exp eriment These detectors enable the annihilation vertex to b e lo cated as well

as the decayvertex if K decays to charged particles To calculate the decayeigentime a

precision of the order of a few millimetres in the vertex p ositions is required but only in the

transverse plane p erp endicular to the b eam axis since it can b e calculated from

M

d

T

p

T

where d is the distance b etween the twovertices p the momentum of the K b oth pro jected

T T

onto this plane and M is the neutral kaon mass

The tracking detectors are followed by the particle identication detector PID which

carries out the charged kaon identication The PID comprises a threshold Cherenkov detector

which is mainly eective for K separation ab ove MeVc momentum and scintillators

which measure the energy loss dE dx and the time of ightofcharged particles The PID is

also used to separate electrons from pions b elow MeVc

The outermost detector is a leadgas sampling calorimeter used to detect the photons

pro duced in decays The design criteria of the calorimeter are mainly dictated by the required

accuracy on the reconstruction of the K or decayvertex The calorimeter provides

e separation at higher momenta p MeVc and is complementary to the PID

The small value of the branching ratio for reaction and the necessary high annihilation

rate place stringent requirements on the exp eriment To reduce the dead time due to data

acquisition and to limit the amoun t of recorded data the unwanted events need to b e removed

eciently A set of hardwired pro cessors hereinafter referred to as HWP has b een sp ecially

designed to achieve this task The role of the pro cessors is to provide full event reconstruction

in a few microseconds chargedtrack patternrecognition and kinematics particle identication

and shower counting in the calorimeter with sucient precision to allowevent selection

Thus the trigger rate can b e kept well b elowkHz

The instrumentation to readout the detectors is based on VME and FASTBUS standards

with the exception of b eam monitoring and slowcontrol which use the G standard

In this pap er wepresent a detailed description of each sub detector used in the CPLEAR

exp eriment The b eam and target are describ ed in Section Section is devoted to the tracking

devices and the global p erformance of these detectors for the charged particle reconstruction

and event kinematics and vertex reconstruction The particle identication detector is discussed

in Section The design and p erformance of the electromagnetic calorimeter are presented in

Section and the trigger pro cessors in Section In the presentpaperwe describ e only the

data acquisition system DAQ of the sub detectors up to the Ro ot Readout system RRO

where these individual data are linked to an eventbuilder The overall data acquisition system

is describ ed elsewhere

The Beam and Target

The antiproton b eam and b eam p osition monitoring

The antiprotons in LEAR are slowed down to MeVc co oled to give a momentum

spread pp of and then extracted overaperiodof minutes with a constantintensity

of ab out MHz A relling p erio d of minutes then follows The b eam has horizontal

and vertical emittance of and cmmrad resp ectively The last part of the b eam

transp ort system to the CPLEAR exp eriment comprises horizontal and vertical b ending

magnets dip oles followed by a quadrup ole doublet These comp onents are used to align and

fo cus the b eam onto the target centre

The general arrangement of the b eam monitoring and target comp onents is shown in

Fig Two orthogonal multiwire prop ortional chamb ers MWPCs are used as b eam monitor

BM They measure the b eam shap e in a plane p erp endicular to the geometrical axis of the

CPLEAR detector in order to ensure that the b eam is centred on the target windowwiththe

smallest p ossible sp ot size mm FWHM

The BM is p ositioned mm upstream of the target entrance and is mounted together with

a scintillator b eam counter on the end of the b eam pip e in a lighttight carb onbre enclosure

The b eam leaves the b eam pip e through a m Be window and exits the BM housing through

am aluminizedMylar window p ositioned just in front of the target window An adjustable

Be degrader is incorp orated in the BM housing and is used to reduce thepmomen tum to

MeVc after passage through all the comp onents so that it can b e brought to rest in the gas of the target

Kevlar wall (0.6 mm thick, radius 7 cm) fibre light guides

H (16 bar) 2 Be degrader 10 cm Beam pipe (vacuum) Target

Be window

140 μm Mylar window

Beam counter scintillator light-tight carbon- fibre hood Beam monitor

MWPC

Figure The b eam pip e b eam monitor and target layouts

It is imp ortanttokeep the b eam wellcentred since the HWP pro cessors assume that b eam

stops at the target centre A b eam centring system is used to automatically p osition the b eam

in the centre of the target entrance window Otherwise the b eam might drift due to p ower

supply instabilities in the b eam line comp onents and cause interactions in the target walls The

BM information is used to make small current adjustments to the horizontal and vertical dip ole

b ending magnets in the b eam line However due to a small misalignment of the b eam line with

resp ect to the magnetic eld axis the horizontal and vertical displacements of the b eam are

coupled A correction matrix is therefore used to give the currentchanges

The b eam MWPC

The MWPCs contain m diameter ano de wires mm apart with pairs of

catho de strips mgoldonm aluminium dep osited on mkapton at right angles

to the ano de wires with a pitch of mm They have a halfgap of mm and are ushed

with a mixture of CO and Ar The low driftvelo city ensures that the displacement

of the charge by the solenoid magnetic eld remains small

The MWPC catho de strip electronics consists of a channel multiplexed currentto

voltage converter followed by an bit ADC which is read by the G DAQ system to pro duce

horizontal and vertical pro jections of the b eam prole Several scans are done and a number of

histograms are averaged to increase the accuracy The mm uncertainty in the mean p ositions

of the b eam centre is mainly due to digitization inaccuracy

The b eam scintillator

The b eam scintillator BS is p ositioned just in front of the target to signal an incoming

antiproton in the target The time b etween scintillator passage and annihilation has an average

value of ns with a time spread of ns FWHM corresp onding to the stopping distribution

along the b eam axis The BS signal is used as the reference time for all other detectors and starts

the trigger sequence

b

The BS is a BC disc with a diameter of mm and a thickness of mm mounted

c

mm upstream of the target window It is inserted in a p olymethyl methacrylate PMMA

d

ring read by lightguide bres glued to the outer circumference The corona of bres is b ent

into the upstream b eam direction group ed in two lightshielded bundles and brought outside

the solenoid magnet where the bres of the two bundles are read by means of a Philips XP

photomultiplier PM The PM is mounted in a Dyno de Timing Base DTB whichprovides

an analogue output and a NIM output from a singlechannel analyser A separate lowerlevel

discriminator is used for pileup rejection

The hydrogen target

The antiprotons are stopp ed in a bar hydrogengas target forming protonium b efore

annihilation

The target is spherical with an inner diameter of mm In order to minimize the

regeneration eects of the neutral kaons the target wall is kept as thin as p ossible It consists

of an inner gastight sphere comp osed of two hemispherical parts pressed from m Mylar



foils in a mould at C and glued together by an additional m Mylar strip over the seam

At the top of each hemisphere the thickness of the Mylar is m The Mylar is covered with

an outer sphere formed of ab out layers of Kevlar bres impregnated with Araldite This

provides the sphere with the necessary strength to withstand the high pressure with a safety

factor of The total wall thickness is mgcm mm An area with a diameter of

mm is left uncovered for the antiproton b eam en trance window A hydrogen supply tub e is

connected opp osite to this window to allow lling of the target

Radiation damage tests have b een p erformed on the window the most fragile part of the

target but no loss in strength has b een found up to Gy

The target sphere is supp orted at the centre of the exp eriment from the rear downstream

side of the solenoid magnet and adjusted to accept the b eam through the centre of the entrance

window

The slowing down of antiprotons in the combination of detectors windows degrading

materials including the rotatable b eryllium absorb er and hydrogen gas leads to a stopping

distribution wellcentred inside the target along the b eam axis FWHM cm and in the

p erp endicular direction FWHM cm

The Tracking Detectors

The tracking detectors comprise prop ortional chamb ers drift chamb ers and layers

of streamer tub es

The multiwire prop ortional chamb ers PC and PC

The two cylindrical multiwire prop ortional chamb ers PCs were designed to b e the inner

most tracking elements of the CPLEAR detector There are two reasons for having prop ortional

chamb er information delivered rapidly to the trigger rstlytovalidate the numb er of tracks

in the initial stages of the trigger logic and secondly to decide on whether or not a trackis

classied as a primary track ie one which either originates from the annihilation vertex or

from neutralkaon decay at short decay time

The cylindrical chamb ers contain axially placed ano de wires m in diameter and with

a pitch of mm and helical catho de strips with a pitch of mm giving a total of

ano de and catho de channels Table The chamber w alls are made of a sandwich

structure of Rohacell foam and Kapton foil Fig This results in a total surface densityfor

b

Manufactured by Bicron Corp oration Newbury Ohio USA

c

Manufactured as well as the PMMA material of Table byRo hm Darmstadt Germany

d

E EskaExtramanufactured by Mitsubishi Japan

one chamb er of mgcm representing radiation lengths X Two wiresupp ort

rings to counteract the electrostatic repulsive forces of the charged wires are also included in

the design These rings are made of the same material as the walls have a width of mm

and are placed some mm apart All channels together with the high voltage and gas

connections are extracted from one end of the detector The whole prop ortional chamb er system

is supp orted in the detector from only the downstream end of the drift chamb er DC structure

The chosen cell geometry identical in b oth chambers is shown in Fig and detailed

information is given in Table These values together with the choice of gas mixture argon

isobutane freon B prop ortions byvolume were derived from an extensive

study

Outer-Shielding (aluminized Kapton-foil) Rohacell Outer-Cathode (aluminized Kapton-foil)

Gas Mixture S

Anode-Wires

Gas Mixture d l Inner-Cathode (aluminized Kapton-foil)

Rohacell Inner-Shielding (aluminized Kapton-foil)

Figure Section through part of a prop ortional chamb er not to scale

Table Prop ortional chamb er parameters see Fig

MWPC Parameters PC PC

Outer radius of outer wall mm

Radius of outer strips mm

Radius of wire plane mm

Radius of inner strips mm

Inner radius of inner wall mm

Active length mm

Wall thickness mm

Full gasgap l mm

Wire spacing s mm

Numb er of wires

Wire diameter d m

Strip width mm

Strip gap mm

Numb er of outer strips

Numb er of inner strips

Strip angle wrt wires deg

Covered solidangle sr

Surface densityofchamb er mgcm

Radiation length equivalentX

The chamb er signal pro cessing is done in two stages Fig the rst incorp orating

the preamplication and discrimination stages is lo cated m away from the chamb ers This

electronics is identical to that used for the drift chamb ers which will b e describ ed in section

It is connected by ns delay cables to the second stage of the frontend electronics

consisting of custom built cluster pro cessors CP channels each and cluster pro cessor

interfaces CPI p er chamb er in FASTBUS crates These mo dules are largely based on

reprogrammable gatearrays XilinxXC and AlteraEP chips The gate congurations

of the arrays in the Xilinx are determined via software whichisdownloaded from a p ersonal

computer via the CPI These interfaces are linked to the trigger system and the RRO system

The input wirehit pattern for eachchamberislatched into a register byaninternally

generated gate signal from the CPI which is initiated by an external trigger signal The

masking clustering and sectoring functions are then executed within ab out ns Wiremasking

can b e used to suppress any hot channels In the clustering function adjacent wirehits are

group ed into single clusters allowing for one eventual missing hit and their clustercentres

are determined In the sectoring stage clusters are asso ciated with a given tracking chamber

element one of minisectors in r This coarse segmented minisector information is sent

to the trigger from where a request can b e sentback for detailed clustercentre information

on a full resolution wirelevel Dep ending on the validityofanevent after full kinematical

reconstruction a strob e signal is sentby the trigger causing the wire addresses to b e transferred to the Ro ot Readout system

Preamp. DATA + FASTBUS Discriminators Detailed Info. CP CPI Detail Info. PC1 Sector VME RRO

trig. TRIGGER FASTBUS control PROCESSORS PC2 Sector Coarse info. Personal Preamp. CP + CPI Detailed Info. Computer Discriminators DATA CPI300

CP300 To CPI300 Simulated trigger OR's Gates Efficiency + Timing Measurement

Simulated Data N I M

Figure Blo ck diagram of prop ortional chamb er readout and monitoring electronics

A system based on fast NIMelectronics is used for testing the whole readout chain as

well as adjusting the timing of the internally generated gate signal used for latching the raw

wire hits into the CP memory

The p erformance gures quoted b elow are based on a nominal highvoltage of V and a

discriminator threshold equivalenttoA This voltage corresp onds to a charge multiplication

factor of As can b e seen from Fig the total currentdrawn by one chamber at

a xed gain or highvoltage is a linear function of the charged particle rate traversing the

chamb er This linearity shows that no spacecharge saturation eects o ccur and has b een tested

for antiproton annihilation rates in the target of up to MHz corresp onding to a charged

particle rate traversing the chamb ers of MHz This is times higher than for the normal

running conditions of the exp eriment

The maximum obtainable eciency of these chamb ers is which is due to the two wire

supp ort rings The measured eciency averaged over the whole chamb er is ab out for the

nominal voltage and discriminator threshold used This leads to a prop ortional chamb er eciency

of at the triggerlevel where a logicalOR is required b etween the twochamb ers The

full plateau eciency characteristics of a prototyp e chamb er measured with electrons from a

collimated Sr source are shown in Fig These characteristics havebeenveried with the nal

chamb ers under b eam conditions but not over the full range of highvoltages The measured

spatial resolution from reconstructed tracks averaged over the whole chamb er is m which

is in go o d agreement with that exp ected from the wire spacing and an average clustersize of wires

5 4.5 2750 V 2700 V A)

μ 4 3.5 2650 V 3 2.5 2600 V 2 1.5 1 2500 V

Chamber Current ( 0.5 0 0 200 400 600 800 1000 1200

Antiproton Annihilation Rate (kHz)

Figure Plot of chamb er currentversus antiproton annihilation rate taken for dierent high

voltage values The plots show no spacecharge saturation eects even at high rates

100

80

60

40 Efficiency (%) 20

0 2000 2200 2400 2600 2800 3000 3200

High Voltage (V)

Figure Eciency curve measured with a prototyp e chamb er and a Sr source

The drift chamb ers DC to DC

Six cylindrical drift chamb ers DCs give the main tracking information The kine

matic cuts calculated online by the trigger pro cessors require a momentum resolution of

This is achieved by combining the measurementpoints in the PCs with those in the DCs six

equidistant p oints spread over a radial distance of cm with a transverse spatial resolution of

m Because of the high eventrates the driftcell size is limited to mm drift time

ns in order to extract the transverse tracking information for the HWP in s Since

the delay constraints do not allow the leftrightambiguitytobesolved using software recon

struction techniques a double sensewire technique separation m is used The drift time

is digitized by a MHz clo ckcounter giving a timing resolution ns

The ano de wires are parallel to the b eam axis and the magnetic eld Fig The elec

trostatic repulsion b etween wires at the same p otential is overcome by gluing spacers made of

small ep oxy balls diameter mm p ositioned every cm along the wires Fig

Mylar skins Field shaping wire Cathode strips

Beam Sense-wire doublet Glue spacers

Total length 2.34 m

Figure Schematics of a drift chamb er showing the sensewire doublet and the helicoidal strips

a)

50 nm Au strips Field shaping Mylar skin, thickness 50 μm angle 11¡Ð26¡ wires at 0 V, ¯ 100 μm

Separation 500 μm

Sense-wire doublet Drift distance at 2.4 kV, 5 mm ¯ 25 μm

b)

20 cm spacing

500 μm spacing

25 μm wire

Glue ball

Figure Detail of the drift chamb er geometry a drift cell b sensewire doublet

To minimize the amount of material in the tracking region the catho de planes consist of

mthick Mylar membranes onto which nm thick gold strips are dep osited byevap oration

The strips are inclined at stereo angles with resp ect to the ano de wires Table to allowthe

measurement of the z co ordinate oine by calculating the centroid of the strip charges

Table Parameters of the drift chambers

Radius Number of Numb er of strips Angle of Numb er of strips Angle of

doublets inner catho de inner strips outer catho de outer strips

cm deg deg

DC

DC

DC

DC

DC

DC

The gap b etween ano des and catho des is maintained by exerting a dierential overpressure

between two adjacentvolumes as shown in Fig The pressure increases from the outer volume

to the inner one which is separated from atmospheric pressure by a cylindrical wall of cm

Rohacell foam of densitykgm The gas mixture used is ethane and argon by

volume

Carbon fibre cylinder +1 mbar Atmospheric pressure +5 mbar +4 mbar DC6

+13 mbar DC5

+24 mbar +25 mbar

Atmospheric pressure

Rohacell wall DC1

Figure Pressure distribution in the drift chamb ers The values shown indicate the overpressure

relative to atmospheric pressure

The main diculty with this typ e of mechanical structure is to maintain a constant gap

to within m rms b etween the ano de wires and the two catho de strip planes Sp ecial

to ols and techniques were develop ed during construction to monitor this gap by measuring the

reection of a ne LASER b eam from the gold strips on the catho de cylinders Cylinders which

did not full the required conditions were rejected The tension of all the wires ab out ton

and the stretched catho de membranes are supp orted by an external rigid cylinder made of a

sandwich of aluminium honeycombbetween two mm thick cylinders of comp osite material

carb on bre

The total material traversed by the charged particles in the drift chamb ers is equivalent

to X not taking into account the outer aluminium and carb on bre cylinder whichis

outside the momentum measurementvolume

For b oth wires and strips sp ecic lownoise preampliers PA have b een develop ed

noise

A for the currenttyp e PA for wires and fC for the chargetyp e PA for strips For the

wires the information is digitized in less than ns using custommade fast TDCs Fig

and is interfaced with b oth the Ro ot Readout system and the trigger system

Root Read-out Wire Preamplifier Discriminator (L or R) L Drift time wire address Left wire

Doublet Trigger TDC processor Right wire R

Gate Clock Threshold

First level trigger Gate 100 MHz

(120 ns) clock

Figure Blo ck diagram of the frontend electronics for the DC wires

The trigger pro cessors receive the following information from the DCs

a bit pattern of the hit wires necessary at the early decision level to select highp kaons

T

coarse segmented minisector information where a minisector represents of the

chamb er circumference used by the track online pattern recognition

on request detailed information in a minisector wire numb er drift time This informa

tion is calibrated using lo okup tables to give the azimuthal co ordinates of the transverse

track which are used to compute the track transverse momentum

The p erformance quoted b elow corresp onds to a voltage of V applied to the sense wires

The ineciency of the wires with a threshold corresp onding to a A input signal to

PA is of the order of Fig mainly due to the use of the spacers b etween adjacent sense

wires There is also a drop in eciency in a region m FWHM around the doublet due to

the low electric eld b etween these wires which are at the same p otential

a

The r resolution in the transverse plane is m Fig but a deterioration of

this resolution for tracks crossing the chamb er can b e seen close to the sense wire at the centre

of the drift cell This is due to the distortion of the electric eld around the doublet mentioned

ab ove which pro duces a lower global amplication As a consequence a signicant time slewing

of the order of ns increases the drift time in this region which aects the error in the

spatial measurement This eect could b e reduced together with the ineciencyby increasing

the voltage in the chamb ers and hence the gas amplication However increasing the voltage is

rendered imp ossible since some antiprotons of the b eam halo annihilate on nuclei of the target

harge of the chamb ers at wall releasing stronglyionizing nuclear fragments which induce a disc

higher electric elds

a

We use a cylindrical co ordinate system with the z axis dened along the incoming antiproton b eam 32 28 24 20 16

12 Inefficiency (%) 8 4 0 Ð0.5 Ð0.4 Ð0.3 Ð0.2 Ð0.10 0.1 0.2 0.3 0.4 0.5

Distance from cell center (cm) Figure Wire ineciency across the drift cell

1000

800 m) μ 600

400 Resolution (

200

0 Ð0.4 Ð0.20 0.2 0.4

Distance from cell center (cm)

Figure Spatial resolution from DC wires across the drift cell

The strips of only four of the six drift chamb ers DC DC DC and DC are equipp ed

with readout asso ciated electronics giving a total of channels The strip charges are not

used in the trigger decision Fig and are therefore only digitized using a standard ADC

in s if the nal trigger conditions are satised Only those strips with a charge ab ove

a threshold of fC after p edestal subtraction are used This pro duces clusters of ab out strips

and a mean charge of fC

The strip eciency is ab out lower Fig than the wire eciency b ecause the max

imum charge induced on the strips is only one third of that induced on the ano de wire The

ineciency intro duced by the presence of spacers every cm is also clearly visible in Fig Preamplifier Line receiver Integrator S/H Analog Strip k multiplexer Subtractor Comparator

Root Read-out (12 bits) Strip n ADC ALU

Gate Pedestal Pedestal (first level trigger) memory

Fast reset (abort trigger)

Threshold Threshold

memory

Figure Blo ck diagram of the frontend electronics for the DC strips

100

80

60

40 Efficiency (%)

20

0 Ð80 Ð60 Ð40 Ð200 20 40 60 80

z (cm)

Figure Strip eciency as a function of the p osition along the wires

The z co ordinate resolution obtained oine by calculating the centroid of the strip charges

is b etter than mm

The drift chambers have b een running since corresp onding to ab out days of

continuous data taking without any ma jor problems and with no evidence of any deterioration

in p erformance due to ageing

The streamer tub e detector

The principal use of the streamer tub e ST detector is to provide a fast readout of the

longitudinal p osition z of tracks to the HWP for use in the online calculation of kinematic

information The detector consists of twolayers of tub es each m long and situated in

the outermost section of the tracking detectors b etween radii of and cm A total of

discrete cm crosssection tub es with a wall thickness of mm and made

e

of extruded MgAl alloy are arranged in identical mo dules each with staggered tub es

e

To BS formely HT containing Mg GallowayTub e Kirkby UK

Fig to coincide with the sectors of the other detectors and to give the b est hermeticity

whilst allowing access for the gas supply

Outer Layer

GAS B A B A B A

CDC DDC

Inner Layer 1.65 cm

Figure Layout of a streamer tub e sector A set of tub es lab elled with the same letter

ABCorDismultiplexed to the same time digitizer

Each tub e has a central ano de wire The dierence of the propagation time of the pulse

along the wire to each end is used to determine the axial track p osition The axial p osition

is used by the HWP to calculate the axial momentum p using the transverse momentum p

z T

obtained from the DCs and assuming that the track is pro duced at the detector centre

The STs are op erated in the limited streamer mo de characterized by large amplitude

voltage pulses having fast rise times Use of the m diameter goldplated tungstenrhenium

ano de wire enables a large gain to b e obtained while maintaining low wire resistance to give

almost equal magnitude voltage pulses at b oth ends of the tub e Thus it is unnecessary to have

active electronics attached to b oth ends of the mo dule

One channel of the streamer tub e electronics is shown in Fig The highvoltage feed

resistor was chosen to allow extraction of pC charge at each end with a voltage drop of less

than V even at the high singles rate in each tub e The circuit provides matching of the voltage

output pulses via characteristic imp edance and mm diameter coaxial cables to the front

end electronics It also has the facility for injection of test pulses to adjust the timetodigital

conversion TDC The pulse from one end of the tub e is delayed by a xed time equal to twice

the maximum transit time along the tub e so that it will always arrive after the pulse from the

other end The pulse which arrives rst starts the timing pro cess and the later pulse stops the

pro cess

The startstop pulses which are typically mV in amplitude with a ns rise time are

conveyed to leadingedge triggered discriminators in the frontend electronics which are set to

a mV threshold to reject noise The output is a ns wide dierential ECL pulse whichis

delayed for triggertiming purp oses by a further m of delay cable b efore en tering the TDC

The fast ECL startstop pulses from three alternate tub es as shown in Fig are pro

cessed into a logical hit for each tub e and taking advantage of the low rate and high conversion

sp eed are connected to the channel multiplexed TDC measuring system based on an bit

ash ADC Thomson TS Fig A customized gatearray pro cesses the start and stop

pulses to check whether they originate from the same tub e and forms a map of valid hits The

TDC has a sensitivity of bit p er ps and the dynamic range of a tub e is bits ns

The TDC is gated by the signal from the trigger Early Decision Logic EDL

The total time b etween the passage of a track and the p oint where the TDC information

is available in the HWP is ns made up principally of drift time ns delay cable

ns and TDC and gatearray conversion time ns A further ns is required for

conversion of the TDC value to a z co ordinate by the HWP

The TDCs which pro cess the hits in one complete sector are interfaced with the Ro ot

Readout system and the trigger HWPTheentire instrumentation system is tested and cali

brated using a test pulse system which applies pulses to the ends of the tub es to create pseudo

start and stop pulses just outside the active length of each tub e Streamer Tube

Delay Start Signal Stop Signal

St VME Root Read-out Discriminator TDC HWP Retrieval Card (3 Tubes) Test Pulse Sp Trigger Gate 330pF Signal 200Ω Wire

0.5ΜΩ H.V.

Passive Electronics

Figure Blo ck diagram of the streamer tub e electronics

RESET GATE

A1 CHANNELS TIMING A2 ORED PULSEWIDTH ADC GENERATOR A3 TOGETHER

STARTS OUTPUT AND STOPS

GATE CONSTANT ARRAY CURRENT LOGIC DISCHARGE

HITS

Figure Detail of the streamer tub e TDC

The gas mixture of argon isobutane and methylal used was optimized togive good

z resolution This is achieved principally through a compromise b etween having a small isobutane

p ercentage to allow streamer development in the gas and a high voltage on the ano de wire to

give a fast pulse rise time The methylal prevents ano de wire dep osits due to p olymerization

of isobutane and therefore aects the lifetime in the highrate environment The most stable

long term conditions were obtained with a mixture of isobutane argon and methylal

and resp ectively and kV ano de voltage giving a wide plateau for the singles rate and a

virtually constant z resolution over the ma jor part of the wire length

The presence of afterpulses from photoninduced secondary streamers can lead to a sig

nicant level of eventoverlap at the highest b eam rates giving rise to false z values from earlier

events b eing transmitted to the HWP The afterpulses are signicantly reduced by adding a

small amount of freon B with little eect on the eciency and leaving the z resolution

unaltered The level of adopted leads to a reduction in the eventoverlap by a factor of

The nal eciency of a single layer of the streamer tub es is over an op erating

voltage range of kV Fig a eciency is lost due to the walls b etween tub es

The z resolution of the streamer tub es is cm for an op erating voltage ab ove kV

Fig b This value is small compared to the width of the stopping distribution of the an

tiprotons in the target which b ecomes the dominan t source of error in p calculated within the

z

HWP

Figure Streamer tub e single layer eciency a and resolution b as a function of high voltage

The ST detector has op erated reliably for years with no sign of long term deterioration

of p erformance Wires taken from tub es after years showed no dep osits and there was no

deterioration of the inner tub e surfaces

Global tracking p erformance

The oine track nding algorithm requires a minimum numb er of out of a p ossible

hits in the chamb ers and has an eciency b etter than p er track

The wire p ositions are determined with an accuracy of m in the transverse plane using

mono energetic pions from the reaction pp

The transverse momentum resolution achieved in the detector is shown in Fig for

two cases dep ending on whether the track is pro duced inside or outside the prop ortional cham

b ers In the rst case the multiple scattering in the material of the tracking detectors contributes

signicantly to the momentum resolution b elow MeVc In the second case the spatial res

olution of the tracking chamb ers is the dominant factor

The missing mass to the primary K pair shown in Fig a has a resolution of the

tted Gaussian of GeVc In the case of the K channel the invariantmass

of the twodecay pions shown in Fig b has a resolution of MeVc

The uncertainty in the distance b etween the annihilation and decayvertices in the trans

verse plane is shown in Fig as a function of the decayvertex radius The resolution of the

decayvertex p osition dep ends on the distance from the vertex to the rst chamb er with hits

and therefore on the radial p osition of the decayvertex but is always within mm Therefore

the error in the K decay time reconstruction is dominated by the K transverse momentum

error as can b e seen in Fig K mean life

S S

(MeV/c)

Figure Relative resolution of the transverse momentum for charged tracks originating from

either inside or outside the PC chamb ers obtained byMonte Carlo simulated data

2 2 σ(M ) = 0.082 (GeV/c 2 ) Number of events

Square of the missing mass to Kπ (GeV/c2)2

σ(M) = 13.6 MeV/c2 Number of events

ππ invariant mass (MeV/c2)

Figure a Square of the missing mass to the primary K pair for selected pp KK events

and b Invariantmassof pair for K decays y ulated ulated data y in the neutral t ertex and the K te Carlo sim yMon te Carlo sim yMon een the annihilation and deca w bet een the annihilation v T w d bet T d tributions due to the uncertain ertex radius obtained b yv erse distance and in the distance T p w separately the con tum time resolution b efore constrained t obtained b es sho y eigen o curv w erse momen ertex as a function of the deca yv ertices aon transv data The t v Figure Deca Figure Resolution in the transv k deca

T contribution fromp resolution decay resolution contribution fromdistance T (K ) 0

However since the exp eriment studies exclusivechannels the resolution in the trackpa

rameters can b e drastically improved by applying kinematically andor geometrically constrained

ts Figure illustrates the resolution obtained for the neutral kaon momentum b efore and

after a constrained t for K events In this particular case a C t has b een applied

with kinematical constraints which require that there is conservation of energy and momen

tum and that the missing mass at the annihilation vertex is the K mass and geometrical

constraints which require the collinearity of the K momentum with the line joining the ver

tices and also require the intersection of the track helices at eachvertex The K momentum

resolution p p improves from to The decay time resolution is therefore im

T T

proved as can b e seen from Fig where it is shown after a constrained t for three dierent

K decaychannels

a) Number of events

Δ b) pT/pT Number of events

Δ

pT/pT

Figure Distribution of p p where p is the dierence b etween the Monte Carlo gen

T T T

erated and measured K transverse momentum p a b efore and b after constrained t see

T

text

Another imp ortant b enet of the constrained ts is that they correct for an overall mo

mentum scale error which could result from imp erfect knowledge of the magnetic eld strength

or from chamb er misalignment Both eects have b een estimated by simulation A shift in

the value of the magnetic eld pro duces a systematic shift in K decay timeof

after the C t for K events and after the C t for semileptonic

decayevents

decay eigentime

Figure Decay eigentime resolution b efore and after N constrained t obtained by

Monte Carlo simulated data The latter corresp ond to dierentK decaychannels

The Particle Identication Detector PID

The Particle Identication Detector PID is used for charged kaon identication b oth

online and oine and for e separation oine A more detailed description can b e found in

The PID consists of trap ezoidal sectors each of which is a liquid threshold Cherenkov

detector sandwiched b etween two plastic scintillators see Fig lo cated b etween radii of

and cm It is supp orted by a structure made of mm thickpointwelded stainless steel foils

S2

Mylar Shroud C Spacers 12.4 cm

S1

Figure Crosssection of one of the PID sectors S S scintillators C Cherenkov detec

tor

which is in turn lying on the calorimeter frame The thickness of each sector is cm

cm for the inner scintillator S cm for the Cherenkov detector and cm for the outer

scintillator S The length of each sector is cm excluding the lightguides and the width

at the half depth is cm The scintillators and the Cherenkov detector are equipp ed with pho

tomultipliers Ateach end the scintillators have one twoinch diameter PM while the Cherenkov

has two threeinch diameter PMs for improved light guide crosssection matching and light col

lection eciency The PMs of the Cherenkov and S detectors are mounted p erp endicular to

the light guide due to the external physical space constraints of the CPLEAR detector see

Fig The PMs of the Cherenkov Philips XPPA were chosen for their go o d single

photoelectron resolution The Cherenkovvessel is constructed by extrusion of UV transparent

PMMA It has a wall thickness of mm and is closed at each end by light guides of the same ma

f

terial This vessel is lled with the liquid radiator C F Fluorinert FC The physical

characteristics of the PID are summarized in Table

Photomultipliers

S2 C Iron C & μ-metal shielding

S1

Light guides

Figure The photomultiplier layout for one of the ends of a PID sector

Table The PID physical parameters

Element S Cherenkov S

Detector

Material BC Extruded PMMA NE A

and radiator FC C F

Width cm

Thickness cm ofFC

Length cm

Light guide

Material PMMA Plexiglas

Length cm

Photomultiplier

Typ e XPB XP XPB

The Cherenkov detector

In order not to pro duce any Cherenkov lightbykaons of reaction the Cherenkov radi

ator should ideally have a threshold indep endent of the wavelength of corresp onding

f

Trademark of M St Paul MN USA

to the maximum charged kaon velo city When energy loss within the liquid radiator and the

inner scintillator is taken into account this threshold is reduced to FC which has a



refractive index of at C in the visible region of the sp ectrum and very low disp ersion

and go o d transparency down to wavelengths of nm is the only liquid available in sucient

quantities which comes close to these requirements

A sp ecic problem is that air has a very high solubilityinFC A dedicated system based

on the use of ultrasound was develop ed to achieve the necessary degassing

The photoelectron yield of the radiator has b een improved by the addition of mgl of the

wavelength shifter diphenyl Oxazole PPO which absorbs lightbetween the wavelengths

of and nm and reemits in the range nm This high concentration of PPO

was also achieved by ultrasound techniques Measurements p erformed with pure FC as the

radiator show that ab out half of the detected photoelectrons are due to the UV comp onentof

the Cherenkovlight which has b een reemitted bythePPO

As the kaons originate from the detector centre there is a strong correlation b etween the

impact p oint along the Cherenkov detector and the incidence angle As a consequence Cherenkov

light cannot b e extracted by total internal reection from the central region of the detector For

this reason a cm region around the centre of the vessel is painted with a white reective

g

paint NE to act as an isotropic diuser This results in a doubling of the photoelectron

yield for tracks traversing this central region signicantly improving the Cherenkov light yield

so that for pions with the photoelectron yield p er PM tub e is everywhere

resulting in a total of more than photoelectrons p er pion

The scintillators S and S

The twoscintillators provide the trigger signals to ensure that a charged kaon candidate

traverses the Cherenkov detector Both scintillators are laterally cm smaller than the Cherenkov

detector in suchaway that tracks emerging from the centre and crossing S and S always pass

through the Cherenkovs sensitivevolume

In addition to their role as a trigger ho doscop e the S and S signals are used b oth oine

and online to identify particles by their energy loss dE dx and time of ight TOF S only

The dE dx measurement is obtained from the digitized signals of each PM corrected

for p edestal and gain In order to b e indep endent of the longitudinal impact p oint and of the

attenuation length the ionization loss is evaluated using the square ro ot of the pro duct of the

signals measured at each end of the scintillator A correction is applied to takeinto accountthe

path length of the particle in the scintillator

To optimize their timing characteristics the S counters are designed to have a go o d light

yield and a geometry which allows go o d detection of nonreected photons thus minimizing

the jitter due to dierent light paths The timing data from each end of S are added to

gether to eliminate the propagation time in the scin tillator giving an intrinsic time resolution of

ps

The timeofight measurementisevaluated as the TOF dierence b etween two tracks in

order to avoid the uncertainty in the annihilation time The latter is only known with a ns

accuracy from the b eam counter signal due to the large uctuations in thep stopping time The

measured time dierence is compared to the theoretical time dierence using a particular mass

hyp othesis and taking into account the ight path lengths from reconstructed tracks The eect

of momentum and ight length resolutions is small compared to the intrinsic time resolutions

stated ab ove

The PID electronics

The PID frontend electronics consists of three customized parts digitization of the pulse

height information digitization of the timing information and a crate controller whichprovides

the interface with the trigger and Ro ot Readout systems

g

Manufactured by Nuclear Enterprises Ltd Edinburgh UK

The Cherenkov and scintillator PM signals are chargeintegrated within a ns gate and

automatically digitized at the end of the gate by bit ash ADCs Thomson TS which

have a digitization time of less than ns For the S and S signals linear digitization is used

Due to the single photoelectron resolution requirements for go o d K separation and the large

dynamic range required for e separation the Cherenkov PM signals are digitized by means of

bilinear digitization

The S signals are pro cessed using constantfraction discriminators CFD to obtain pulse

height indep endent timing information The time measured is that b etween the b eam counter

signal and the S signals A custommade analog circuit has b een designed to stretch the typ

ical timing of ns to ns to obtain b etter resolution Digitization is p erformed in

ns byDATEL ADCs DATEL ADC a bit ash ADC followed by a bit successive

approximation ADC

The custommade controller is built to handle requests for ADC and TDC data from the

trigger at a rate of MHz and to handle readout rates of Mbytess Fast ECL logic is used

in the trigger interface section and TTL logic in the readout stream To reduce the readout

time zero suppression has b een implemented by the use of hit maps which are logical ORs of

the S C and S signals

Calibration

A realtime calibration pro cedure for the PID detector is needed since the PID information

is used by the online trigger pro cessors and provides most of the trigger rejection Most of the

calibration constants are measured using pionic tracks from minimumbias events Because of

the go o d stability of the PID detector and its electronics it is only necessary to p erform the

calibration every hours The p erformance of the PID detector is p ermanently monitored

using a subsample of physics events which is fully reconstructed and analysed by a dedicated

workstation The calibration constants consist of

the p edestals and minimum ionizingparticle p eak p ositions after correction for attenua

tion length for eachPMinthescintillators S and S

the p edestals and single photoelectron p eak p ositions as well as the relative yield of the

v arious PMs for the Cherenkov detector The latter is obtained by measuring the mean

numb er of photoelectrons detected in each PM for energetic pions crossing the central

region of the PID detector

the individual delays of each PM for the TOF measurement in S

The calibration constants are made available to the online pro cessors by means of lo okup tables

which translate the raw information into physical quantities

PID p erformance

The PID data are used online at two stages of the trigger see Section the Early Decision

Logic EDL based on the threshold Cherenkov resp onse and a transverse momentum cut

denes a kaon candidate in ns the hardwired pro cessor HWP renes the kaon identication

after s using energy loss and timeofight information from the scintillators

At the EDL level a kaon is dened by the coincidence of the twoscintillator layers and

the absence of any Cherenkov resp onse S C S in any sector of the PID The C signal is

fullled if not more than one PM has a signal greater than photoelectrons This was chosen

to b e a go o d compromise b etween a more stringent requirement whichwould have minimized

ineciencies in the counter and a more relaxed one whichwould not have removed those kaons

either pro ducing accidental background light Cherenkov light pro duced in the plastic b ox and

light from kno ckon electrons or slightly ab ove the Cherenk ov threshold The eciency of the

S C S criterium has b een measured at for kaons with a momentum greater than

MeVc This measurementwas achieved with a sample of kaons free from any trigger bias

taken from pp K K events selected using kinematically constrained ts The measured

eciency is shown in Fig as a function of the kaon momentum Atmomenta b elowMeVc

the probability that a kaon pro duces an S C S signature drops This is b ecause these low

momentum kaons do not reach S since their energy loss and interaction probability in S and Cishigher

1

0.9

0.8

0.7

0.6

0.5 Efficiency 0.4

0.3

0.2

0.1

0 100 200 300 400 500 600 700

Kaon momentum (MeV/c)

Figure Kaon identication eciency as a function of momentum measured from pp

K K events

1

-1 10

-2 10 Probability of pion misidentification 5.7 × 10Ð3 →

100 200 300 400 500 600 700 800 900

Pion momentum (MeV/c)

Figure Probability of pion misidentication as kaon at the EDL level

The pion rejection eciency can b e evaluated using welldened pp events

selected by kinematically constrained ts Monte Carlo studies show that the contamination

of this sample bykaons is less than The S C S criterium as a function of

pion momentum is shown in Fig Atlowmomenta b elow MeVc the pion rejection is

inecient due to those pions which are b elow or just ab ove the threshold or which are strongly

interacting in the detector material These low momentum pions that could fakeakaon signature

are removed at the EDL level by a transverse momentum p cut and during oine analysis

T

by momentum cuts This p cut removes only a small fraction of the charged kaons

T

pro duced in reaction For pions ab ove MeVc the mean probability of faking an S C S

signature has b een measured to b e less than at EDL level using the previous sample

Fig It should b e noted that this result is not explained by statistical ineciencies whichare

calculated assuming P oisson statistics of the photoelectrons in the Cherenkov The ineciency

is mainly due to pions interacting inside the Cherenkov or the scintillator material and leaving

aS C S signature This explains why the pion rejection has remained constant despite an

approximate p er year light decrease in the Cherenkov during the years of op eration due

to FC p ollution by plastics

Figure shows the dE dx value as a function of momentum for kaons and pions in

scintillator S The resolution is b etter than FWHM allowing go o d K separation up

to MeVc The S counter b eing thinner has a p o orer resolution of The resolution

in the TOF dierence b etween the twocharged tracks from the annihilation vertex is ps

Fig allowing a separation of between K and pair hyp otheses for kaon transverse

momenta up to MeVc and up to MeVc

Figure Energy loss measured in S as a function of momentum For b oth kaons and pions

solid line shows the exp ected average energy loss and the dashed lines form a band containing

of the events

The PID information is also used to p erform e separation using neural network tech

niques in the momentum range MeVc Within the momentum range MeVc

an eciency of is obtained for electron recognition which decreases to at MeVc

with a constant pion background of see Fig 1400

1200

1000

Kπ hypothesis 800 σ = 255 ps ππ hypothesis

Number of events 600

400

200

0 Ð3 Ð2 Ð1 0 1 2 3

TimeÐofÐflight difference (ns)

Figure Timeofight dierence b etween K and for the whole accepted phase space of

K K events The solid histogram corresp onds to the correct particle assignment the dashed histogram to the case where b oth particles are assumed to b e pions Electron identification efficiency

Electron momentum (MeV/c)

Figure Electron identication eciency from PID information as a function of the momen

tum

The Electromagnetic Calorimeter ECAL

To study nal states with neutral pions ie K K wehave

develop ed a ray detector ECAL Electromagnetic CALorimeter with a number of novel

features

The ECAL is constructed in the form of a barrel without end caps the solid angle coverage

is The barrel is an assembly of lead plates alternating with sampling chamb ers which

consist of small crosssection tub es and pickup strips The technique chosen p ermits a high

degree of granularity in b oth r randz co ordinates in addition to readout exibilityThe

p

ECAL is characterized by a total of X an energy resolution of E E GeV

and mm p osition resolution for the photon conversion p oints

Design criteria for the ECAL

Due to the photons low energy MeV a go o d photon detection eciency can only

be achieved with the ECAL situated inside the magnet coil and with a ne sampling fraction

Only a mo derate calorimeter thickness in terms of X is required

The design studies have shown that for a K decaying to or the most imp ortant

input in the determination of the decayvertex is the p osition of the conversion p ointinthe

detector A fairly p o or energy resolution can b e accepted provided the spatial resolution on

the conversion p oint is b etter than mm However even if it is not accurate the energy

measurement is useful for background rejection and photon pairing The simulations havealso

shown that at low energy the b est spatial resolution in the conversion p oint is obtained by

sampling the shower as close as p ossible to the conversion p oint As a result the detector

demands high granularityinthe rplane as well as go o d longitudinal segmentation in order to

resolve the conversion p oints of s impinging at a small angle with resp ect to the detector axis

Geometrically the design of the ECAL was constrained from b oth the inner and the outer

radii The K decayvolume instrumented with the tracking devices and surrounded bythe

cm thick PID denes a minimum internal radius of cm for the calorimeter Externally

the ECAL had to t within the magnet coil of m radius

All the previous considerations concerning geometrical constraints and optimization of

p erformance and cost dictated the choice of a leadgas sampling calorimeter rather than a

crystal calorimeter such as CsI

Small crosssection highgain tub es were develop ed as the active medium for the

calorimeter A prototyp e of ECAL reduced in size but for number of X was successfully tested

in a e and b eam and the results have already b een rep orted

Construction of the ECAL



The ECAL consists of three azimuthal sectors each one containing layers identical

in depth Eachlayer equivalentto X contains a converter plate followed by tub es of

mm active length which are sandwiched b etween strip plates Fig The converter is made

of mm lead sandwiched b etween two m thick aluminium foils reinforcing the lead The

mm crosssection tub es op erated near to the limitedstreamer mo de are built in the

form of gastight packages comprising eight separate cells hereinafter referred to as chamb ers

There are chambers per layer and p er sector eachof mm crosssection Both

the top and b ottom covers of the chamb ers are covered with graphite on the gas side and on

the outside they are in close contact with copp er pic kup strip plates The calorimeter contains

a total of chamb ers tub es or wires The chamb ers are op erated at kV with a

gas mixture of CO and p entane byvolume The gas supply is serial through the

chamb ers of a layer The high gain of the tub es p ermits their readout without amplication

The typical pulse from a wire is mV and has a width of ns as measured for a load



The strips are arranged such that they cross the wires at an angle of allowing the

reconstruction of the z co ordinate in eachlayer The strip plates consist of a copp er ground

plane insulated from the copp er strips which are mm wide at a pitch of mm The end of

each strip which is not connected to the readout is grounded via a resistor of equal to the

strip characteristic imp edance The strip signal typically A is fed into a currenttovoltage

amplier mVA in order to give a signal similar to that from the wire There are a total

of strip channels

The total weight of the ECAL is tons which necessitates the mechanical structure

shown in Fig It consists of two mm thick aluminium cylinders connected by mm thick

  

radial aluminium bars at three azimuthal angles to form three separate azimuthal

sectors The load of the tracking devices and the PID is susp ended on the ECAL supp ort

structure In turn the load of this structure is distributed on four mm thick aluminium rings

interleaved with the magnet coils Fig a) U strip V strip Al support 15 mm

0.3 mm Al 9.5 mm 1.5 mm Pb

0.3 mm Al

0.3 mm Al glued on lead In total 18 layers

Al support 15 mm

b)

0.6 mm Copper shield

V strips

30 ¡

G10

Etched printed board

5.6

6.6 30¡ U strips

Figure a Transverse crosssection of the calorimeter b Geometrical setup of the layers

of strips pressed on the top and b ottom covers of a chamb er All lengths are in mm

A a) b) 9.5 2 2.5 °

120 × 3

Al plated glued on lead 201.5

3 spacers

3340 long 18 steps of 7.5 mm

3 20 ° 30° 30 Glued and screwed 30°

Detail A

Figure a Transverse crosssection of the mechanical supp ort structure of the calorimeter b

Detail of the spacer b etween sectors All lengths are in mm

Readout of the ECAL

The challenge of the calorimeter data acquisition system is to provide the information

of channels within a few microseconds for the secondlevel trigger pro cessor and to nish

the complete readout in less than one millisecond Mo dern design techniques and parallel data

transfers havebeenusedtoachieve the required p erformance The Fastbus standard is used throughout Fig

16 CHANNELS 8 CHANNELS AMP WI

o o 4 AMP.-CARDS 24 CRAMs 8 CHAMBERS 24 CRAMs o o AMP WI

CR CRo o o o o CR CR CR o o o o o CR 42 CRATES

oooooooo [FASTBUS] [FASTBUS] FRONT-END CRATE FRONT-END CRATE CP [STRIPS] CP [WIRES]

o o CPT- o . DRIVER O O O O O O O O O O O O O O 42 CHANNELS . 42 CPMs 42 CHANNELS

o o o o o o o o CMCM CC CM CM CC FE FEFE FE

2 CPM HWP2.5 CRATES CRATE [FASTBUS] [CPVME] CF CF

AMP.....Preamplifier CC...... Cal. Trigger Controller (CTC) TO CF...... CERN Host Interface (CHI) and EVENT-BUILDER Fastbus VSB Interface (RRO) CM...... Cal. Pattern Memory (CPM) CP...... Cal. Pattern Transmitter (CPT) CR...... CRAM FE...... Front-End Trigger Proc.

WI...... Wire chamber HV cards

Figure Blo ck diagram of the calorimeter readout

The wire signals are transmitted to the frontend mo dules via twisted pair cables one side

of the pair b eing directly driven from the wire the other terminated via a resistor to

ground Strip signals are transmitted in dierential mo de from the driver of the preamplier A

blo ck diagram of the readout is shown in Fig

The wire and strip signals are received by the frontend mo dules Calorimeter Receiver

And Memory CRAM Each mo dule contains channels of input receivers discriminators

h

shap ers and digital delays Up to mo dules are in one Fastbus crate thus requiring a total of

mo dules distributed in Fastbus crates for the readout of channels The mo dules

run in an autoup dating mo de every ns which makes resetting unnecessary Additional

features are selectable afterpulse cancellation times set to ns and selectable acceptance

windows for the output register set to ns Test pulses can b e sent directly to the HV

b oards and the inputs of the preampliers or the internal output registers This allows a detailed

verication of the calorimeter readout chain and a fast check of its p erformance

The information of up to CRAMs is read via a sp ecial crate controller Calorimeter

Pattern Transmitter CPT using a daisy chain proto col It is transferred as hit group pattern

h

Application Sp ecic Integrated Circuit ASIC

HGP containing the hitmap of chamb ers and as data pattern representing the exact hit

distribution of chamb ers This granularity enables fast online reconstruction of showersinall

three pro jections via an external secondlevel trigger pro cessor HWP see Section

The central distribution system CPTdriver provides signals such as clo ck trigger test

pulses and discriminator thresholds via each CPT to all CRAM mo dules in a crate

All CPT mo dules are connected to corresp onding pro cessing mo dules Calorimeter

Pattern Memory CPM which provide fast data collection send the HGP informationtothe

HWP pro cessor and enco de the data thus minimizing the quantity of data to b e recorded

on tap e An average event of ab out red channels out of channels in total will give

bit words after enco ding A rstlevel decision starts the data transfer of all CPTs

in parallel to s The enco ding of the data takes b etween and s dep ending on

the numb er of data patterns HWP calculates in parallel the numberofshowers seen bythe

calorimeter and decides within s whether the event should b e written to tap e

The nal readout is p erformed in parallel bytwoFastbus master mo dules CERN Host

i

Interface CHI after the event has passed the last trigger decision The readout of mo dules

takes sfora typical event bytes The data memory of each CHI is seen by the event

builder via VSB mo dules A small fraction of the ev ents ab out out of is sent via Ethernet

to a pro cessor to monitor online the p erformance of the calorimeter

Performance of the ECAL

The p erformance of the ECAL the photon detection eciency the photon conversion

p osition resolution the energy resolution and the charged particle identication dep ends for

the oine analysis on the shower pattern recognition

A tailored shower pattern recognition algorithm searches for clusters of hits separately in

the pro jection of the wires W and the two strip planes U V The minimum requirements

for a W pro jection are hitsin layers and for b oth U and V pro jections hits in

layers A shower is dened by the matching of W U and V pro jections in a single layer taking

advantage of the UV redundancy The matching provides the shower conversion co ordinates

ie the photon conversion p oint in the ECAL A shower is dened as a neutral shower if the

shower conversion p oint is separated from anycharged track extrap olated in the magnetic eld

through the PID into the ECAL by at least cm in b oth the rplane and the z direction

The photon detection eciency is measured with tagged photons from the neutral pion

decayofthetwob o dy annihilation reaction

pp K K

The momentum vector is calculated from the K K momentum vector With the informa

tion from one detected photon the conversion co ordinates in the ECAL can b e reconstructed

for the other one If the second photon is reconstructed to b e within the activevolume of the

ECAL the eciency is inferred by counting the events where the second neutral shower is found

in the exp ected region The photon detection eciency as a function of the photon energy is

shown in Fig reaching a plateau of at a photon energy of MeV Ab out of the

ineciency is due to the supp ort structure The ineciency for lowenergy photons is partly due

to the detector material in front of the ECAL which amounts to X for the target and the

tracking devices and to X for the PID

The detection eciency for electrons and p ositrons studied using pairpro duction by conversion

in the detector material is found to b e very similar to photons except that it is shifted by

MeV due to energy loss in the PID

i

H Muller et al CERN Host Interface Users manual CERNEP 1

0.8

ηγ = (90 ± 1)%

0.6

Efficiency 0.4

0.2

0 0 100 200 300 400 500 600

Eγ (MeV)

Figure Photon detection eciency as a function of the photon energy

The spatial resolution for electromagnetic showers is deduced from the photon op ening

angle distribution in the decay of mono energetic MeV s from the reaction given in

The measured distribution has b een tted with the exp ected distribution with a minimum



op ening angle of The overall op ening angle resolution for the two photons is unfolded and

amounts to mrad which translates into a global conversion p oint resolution for one

angle

photon in space of mm

space

The spatial resolution of the calorimeter can also b e studied by comparing the r and

z co ordinates of extrap olated charged tracks with the corresp onding co ordinates in the ECAL

determined byshower pattern recognition A resolution of mm was found in the rplane

after subtracting the contributions of multiple scattering mm and of measurementerror

in the tracking chamb ers mm to the track extrap olation uncertaintyInthe z direction a

precise evaluation is not p ossible b ecause of the large track extrap olation error mm due

to the DC measurement error dominating our resolution measurement

Simulation studies show that in a sampling calorimeter the b est energy resolution for

photons with only a few hundred MeV energy is obtained bycounting the number of track

segments rather than measuring the energy dep osited in the gas

Figure shows the hit multiplicities for tagged photons of a MeV and b

MeV energy showering in the rst layer of the calorimeter The distribution is asymmetric

y the shower pattern for MeV due to the minimal shower requirement of three hits imp osed b

recognition algorithm To parametrize the hit distributions a dedicated function is constructed

to account for the asymmetric shap e neither a Gaussian nor a Poisson function successfully t

the distributions

w w E

H w E N E exp

E E w w jw w j

where w is the numberofhitsand w E is the hit multiplicity for which the function reaches

the maximum value The parameter E takes account of the asymmetricshapeFor photon

energies MeV the hit distribution is symmetric E vanishes and the function b ecomes

a Gaussian with a variance equal to E Atlower energies a go o d approximation is given by

an eectivevariance E E

w a) b) σ σ = 1.99 ± 0.12 30 = 5.84 ± 0.54 300 0 0 Ω = 1.66 ± 0.16 Ω = 0.47 ± 1.62

w0 = 5.81 ± 0.10 w0 = 27.1 ± 0.53 25 250

200 20

150 15 Number of events Number of events

100 10

50 5

0 0 10 20 30 40 50 60 10 20 30 40 50 60

Number of hits Number of hits

Figure Distribution of the numb er of hit tub es in the calorimeter for two photon energies

a MeV and b MeV

In Fig the variation of the mean value wE is displayed as a function of the

photon energy and of the layer in which the photon converted The deviation of the data p oints

from a straight line reects the limited thickness of the calorimeter leading to shower leakages

esp ecially for photons converting in layers further out in a radial direction (hits) > W (E) <

Energy (MeV)

Figure Mean value wE of the hit distribution as a function of the photon energy for

three conversion layers

The resolution of the hit distribution as a function of the photon energy dep ends only

weakly on the conversion layer The reduced resolution R dened as

p

E GeV

w

R

wE

is almost constantover a wide energy range as shown in Fig

R = 0.126 ± 0.004

R = 0.129 ± 0.008

R = 0.175 ± 0.022

Figure Reduced resolution see text as a function of the photon energy for three photon

conversion layers at dierent depths

The describ ed p erformance of the ECAL enables the reconstruction of the K decay length

with an average precision of cm FWHM for the nal state of two neutral pions K

where all the four photons have b een detected This translates into a resolution in the K

decay time of A detailed discussion of the analysis and results can b e found in

S

The distinction b etween electron and pion showers in the calorimeter was exploited for

e separation The ne granularity allows for a precise determination of shower top ologies The

neural network technique was applied for the separation using the ECAL pattern information

for the showers as input and relying exclusively on the wire hits b ecause of their lownoise and

higheciency p erformance Calibration data fromp p annihilation to four or six charged pions

as well as converted photons e e were used to train the network The purity of these

samples was

The ECAL pattern data were pro cessed to yield the relevant information for presenting

the shower characteristics to the neural network The training was p erformed in the momentum

range MeVc with MeVc bins Atlower momenta MeVc pions do not

reach the ECAL thus allowing an obvious separation

With the calibration data an eciency of in identifying electrons was achieved at

pion contamination over the entire momentum range Fig

0.9 0.06 0.8 0.05 0.7

0.6 0.04

0.5 0.03 0.4 Efficiency

0.3 Contamination 0.02 a) electrons b) pions 0.2 0.01 0.1

0 0 200 400 600 200 400 600

p (MeV/c) p (MeV/c)

Figure Electron identication eciency a and pion contamination fraction b as a function

of the momentum using the showers in the calorimeter

The CPLEAR Trigger

The small annihilation branching ratio for the desired channels of K K pro duction

and the high b eam intensities require an intelligent and fast trigger system to select them from

the unwanted background This led to the implementationofamultilevel hardwired pro cessor

system HWP capable of utilizing the detector resp onses as so on as they b ecome available for

the selection pro cess

The trigger system aims at the b est p ossible selection of K and K in identifying a primary

 

K pair and at the observation of the K K decay inside the ducial volume of the detector

Nearly half of the pro duced neutral kaons of the K comp onent decay outside this volume

L

The decisions are based on fast recognition of the charged kaon using the PID hit maps

the numb er and top ology of the charged tracks the particle identication using energyloss

timeofight and Cherenkov light resp onse and kinematic constraintsaswell as the number

of showers in the ECAL

Figure shows the trigger decision steps in chronological order taking into accountthe

availability in real time of the relevant detector information The rate reductions pro duced by

t pro cessor steps given in Fig dep end on the particular detailed logic conditions the dieren

set at each level

The rst decision step Early Decision Logic EDL selects candidate events with at least

twocharged tracks at least hits in the inner scintillator S one of whichmust b e a kaon

dened by the coincidence of the two scintillator layers and no Cherenkov resp onse ie S C S

This rst identication of charged kaons is further improved by requesting a minimum transverse

momentum p cut in order to eliminate false kaon candidates due to slow pions

T

The second step Intermediate Decision Logic IDL identies kaonic and other charged

track candidates using sectorized information hit maps from the tracking chamb ers in con

junction with the PID and classies them dep ending on whether they originate inside primary

track or outside secondary track the prop ortional chambers CPLEAR Fast Trigger

PID sectors ⇒ Early Decision Logic 60 ns, Rate reduction: 4.1 S1, C, S2 N(K)›0, Ntrack›1 Fast Kaon Candidate DC1, DC6 ⇒ 400 ns, Rate reduction: 3.1 wires p T cut ⇒

PC, DC, PID ⇒ Intermediate Decision Logic 80 ns, Rate reduction: 1.7 sectors K, primary, all charged tracks

Track Follower Pattern Recognition PC-, DC-, ST- ⇒ 1.9 μs, Rate reduction: 1.4 wires Track Validation ⇒

Track Parametrization Curvature, φ, z, Distance of min. approach Momentum Calcul.

500 ns, Rate reduction: 1.3 ⇒ p , p T z ⇒

S1 ADCs Missing Mass dE/dx, TOF, ⇒ S1 TDCs p +p = const. N(e) in Cerenkov K π C ADCs Kinematics Particle Ident. 500 ns, Rate reduction: 1.5 1.9μ s, Rate reduction: 6.4 ⇐

ECAL Fast ECAL Processor Shower Calculation ⇒ N(shower)› n wires 0 17 μs, Rate reduction: 2.9

Read-out starts after 34 μs ⇐

Root Read-Out To Tape drive

1.1 ms

CPLEAR

Figure Logic and data ow diagram of the trigger system

The third step HWP pro cessor includes

ne grain tracking and gathering of the frontend electronics digital information from the

tracking chambers TrackFollower and TrackValidation

track parametrization to compute the track curvature the azimuthal angle at the

p oint of minimum approach in the transverse plane the distance of minimum approach

and the derivation of the z co ordinate from the streamer tub es

calculation of transverse and longitudinal momenta of the primary particles and appli

cation of a kinematical cut on a minimum value of the sum p p equivalenttothe

K

missing mass for all K pairs

P

the charge balance Q required for events with an even number of tracks or

i

The quality of the tracknding and parametrization algorithm can b e seen from Fig where

the track curvature computed by the trigger pro cessor HWP is plotted as a function of the

curvature measured oine for events

0.02 )

1

Ð 0.01

0

Curvature (Trigger) (cm -0.01

-0.02

-0.02 -0.01 0 0.01 0.02 Ð1

Curvature (Offline Pattern Recognition) (cm )

Figure Online versus oine track curvature

 

The next step HWP pro cessor improves the K pair identication by comparing the

measured energyloss and Cherenkov resp onse for the kaons as well as the TOF dierence

t t t to their exp ected values according to the previously determined track parameters

K

For candidate neutral events ie events with only two primary K tracks the showers

in the ECAL are quickly identied by a dedicated pro cessor HWP The HWP pro cessor

examines the ECAL hits and determines the number of showers using an algorithm acting on

the hit group pattern information in the rplane This has proved to b e the most p owerful

metho d for the identication of electromagnetic showers given the low energy of the photons

There is no distinction b etween charged and neutral showers The requirement for a minimum

numb er of showers n or eliminates all K events decaying outside the ducial volume

L

of the detector

The overall rejection factor of the trigger is allowing a readout rate of events

erage b eam rate of kHz The global acceptance for K K pro duced in p er second at an av

reaction and decaying in the ducial volume of the detector is mainly determined by

geometrical factors

The decision time of each stage varies of the trigger from ns EDL to s HWP

The readout strob e is delivered to frontend electronics s after b eam counter signal arrival

The trigger pro cessors have b een constructed using custom ECL electronics VME stan

dard An internal system clo ckisusedtosynchronize the data ow and logic decisions

starting with the ne grain tracking in HWP

The communication and data owbetween the trigger pro cessors and the detector elec

tronics b enets from the readout sectorization in the transverse plane ie PID sectors

and tracking chamb er minisectors Data requested by the pro cessors using the appropriate

minisector address are returned from the frontend electronics hardwired pro cessor interfaces

together with the information requested This form of dialogue is faster than using a handshake

pro cedure

Lo okup tables are used whenever the trigger pro cessors request online calibrated front

end data and for the numerical computation of physical quantities in particular for the track

parametrization stage and for the HWP pro cessor

The multilevel pro cessor system is controlled using a pip elined logic sequencer unit A

trigger control system provides all the signals required by the detector frontend electronics and

initiates the transfer of information to the Ro ot Readout system It is capable of rejecting a

wrong candidate event as so on as it is p ossible following its identication After a clear time of

ns the trigger system is ready to accept the next candidate event This minimizes the trigger

deadtime due to the large numb er of rejected events The overall deadtime of the trigger is

ab out at a MHz antiproton rate

The p erformance and stability of the trigger system are monitored online using data gath

ered by a dedicated interface called SPY This collects the data input to the pro cessors the

hit maps and frontend data from the detectors and all intermediate and nal data and logic

decisions generated by the trigger system

By using the SPY data and the results of a full trigger simulation which includes a sim

ulation of the digitization for all pro cessors the functionality and the eciency of eachstage

can b e measured For each stage the functionality is determined by comparing the decision of

the trigger with the decision computed by the trigger simulation using the input data to this

particular stage as read by the SPY The functionality reects the b ehaviour of a particular

trigger stage including internal timing imp erfections For all stages the functionality is b etter

than To determine the eciency of each stage the input information of the stage used by

the simulation is reevaluated using the Ro ot Readout data of each subdetector The eciency

therefore dep ends not only on the functionality but also on the quality and timing imp erfections

of the data transfer b etween the subdetector frontend electronics and the trigger During data

taking the eciencies are measured each hour on a data sample of events and the func

tionalitychecks are p erformed once a day The overall trigger eciency continuously measured

during the run is around with v ariations at a level of

Summary

The CPLEAR detector has b een fully op erational since All its sub detectors

b eam monitors tracking devices particle identication detector electromagnetic calorimeter

haveachieved their exp ected p erformance Sp ecially designed hardwired pro cessors havebeen

successfully implemented to trigger the exp eriment and remove the unwanted background The

detector has run smo othly during the last years and accumulated nearly decays of

strangenesstagged neutral kaons entering our nal data sample allowing precision studies of

CP T and CPTsymmetries

Acknowledgments

Weacknowledge the imp ortantcontribution made by the PPE technical supp ort group

G Muratori and M Price We thank D Fromm and H Rigoni for their help with the gas dis

tribution system and C Millerin and his team for their assistance in the electronics pro duction

This work was supp orted by the following institutions the French CNRSInstitut National

de Physique Nucleaire et de Physique des Particules the French Commissariat a lEnergie Atom

ique the Greek General Secretariat of Research and Technology the Netherlands Foundation

for Fundamental Reserach on Matter FOM the Portuguese JNICT and INIC the Ministry of

Science and Technology of the Republic of Slovenia the Swedish Natural Science Research Coun

cil the Swiss National Science Foundation the UK Particle Physics and Astronomy Research

Council PPARC and the US National Science Foundation

References

E Gabathuler and PPavlop oulos Strong and Weak CP violation at LEAR Pro c

Workshop on Physics at LEAR with Low Energy Co oled Antiprotons Erice eds

U Gastaldi and R Klapisch Plenum New York

D Troster et al Nucl Instrum Metho ds A

CP Bee et al The CPLEAR data acquisition system Pro c CHEP Conference CERN

CP Bee et al The CPLEAR data acquisition system in preparation

M Van den Putte et al IEEE Trans Nucl Sci

M Van den Putte PhD thesis Delft UniversityofTechnology Netherlands

J De Vries PhD thesis Delft UniversityofTechnology Netherlands

RW Hollander et al Materials Science Forum

Ch Bula PhD thesis Swiss Federal Institute of TechnologyZurich Switzerland

M Dejardin et al Nucl Instrum Metho ds A

M Dejardin PhD thesis UniversityofParis VI I France

PM Sanders PhD thesis UniversityofLiverp o ol UK

JM Bennet et al A Streamer Tub e detector for op eration at high rates in the CPLEAR

exp eriment at CERN Nucl Instrum Metho ds in print

DM Lee et al Nucl Instrum Metho ds A

M Do dgson PhD thesis UniversityofLiverp o ol UK

R Rickenbach et al Nucl Instrum Metho ds A

A Angelop oulos et al Nucl Instrum Metho ds A

M Fluorinert Pro duct Information

M Danielsson and HJ Gerb er Nucl Instrum Metho ds A

A Chlopik et al Nucl Instrum Metho ds A

WE Earle et al Nucl Instrum Metho ds A

M Dejardin CPLEAR note CPSW

R Adler et al The CPLEAR electromagnetic calorimeter in preparation

P Blo ch et al Nucl Instrum Metho ds A

R Adler et al Nucl Instrum Metho ds A

F Bal et al Nucl Instrum Metho ds A

B Pagels et al Reconstruction of a particle decaycontaining intermediate decays of invis

ible neutrals ETHZIPP Internal Rep ort

DA Troster et al CPLEAR note CPHWP

R Adler et al MeasurementoftheK K mass dierence using semileptonic decays of

L S

tagged neutral kaons Phys Lett B

R Adler et al Measurement of the CP violation parameter using tagged K and K

Phys Lett B

R Adler et al First observation of a particleantiparticle asymmetry in the decay of neutral

kaons into ZPhys C

R Adler et al Search for CP violation in the decay of neutral kaons to Phys

Lett B