HAL Id: in2p3-00010669
http://hal.in2p3.fr/in2p3-00010669
Submitted on 10 Dec 2001
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
The scintillator option for the OPERA target tracker:
results on the readout
L. Chaussard, P. Jonsson, S. Katsanevas, J. Marteau, G. Moret, S. Gardien,
C. Girerd, D. Essertaize, G. Guillot, S. Vanzetto
To cite this version:
July06,2001
The scintillator option for the OPERA target tracker:
results on the readout.
L.Chaussard, P.Jonsson,S.Katsanevas,J.Marteau, G.Moret
S.Gardien,C.Girerd; D.Essertaize,G.Guillot, S.Vanzetto
InstitutdePhysiqueNucleairedeLyon
UniversiteClaudeBernardLyonI
F-69622Villeurbannecedex
Abstract
Thisnote summarizesthestudiesperformedat theIPNLconcerning thereadoutofthe scintillator
optionofthetargettracker(TT).Inparticularwereviewtheperformancesandeventuallimitations
ofthemultipixelphotodetectorHPD,theautotriggerablelownoiseVikingChips(VATAseries)and
anethernetbasedDACQsystem. Wepresentbeamtests,includingthefullreadoutchainfromthe
photodetectortotheworkstation,withboththeliquidandplasticscintillatoroptions.Basedonthe
abovewepresent analternative tothe baselineoptionfor the placement ofthe photodetectors. In
thisnewarchitecturethephotodetectorsandtheassociatedelectronicsareputatthecornersofthe
walls,farfromtothescintillatormodules.Westudytheimplicationsofthischoice,andpresentwork
inprogress concerning a prototype. We thenreviewthe implications ofthe current electronicsfor
thereadoutand event-buildingarchitectureandpresentresultsonthesimulationofthefullDACQ
ThebaselineoptionoftheOPERAtargettrackerforeseestheuseofplasticscintillator,readbya
multi-pixelphoto-detector(MaPMT)andauto-triggerableelectronics. TestsandMonte-Carlosimulationshave
beenundertakeninthecollaboration,andinLyoninparticular,tocharacterizetheperformanceofvarious
alternatives[1].
Inthisnote,insection2wepresentthetestswehaveperformedwithHybridPhotoDiodes(HPD's),
photodetectors that have recently reached a great level of maturity. We then proceed to present the
characterization of the VATA autotriggerable electronics (section 3). These electronics are suitable,
thoughdierentversions,forboththeHPDandtheMaPMT.Thecurrentperformancesaresummarized
and a set of benchmarks for the LAL/Strasbourg chipunder constructionare implicitly set. Wethen
(section4) presentthe dataacquisition card ORCA, constructedin Lyon,and validatedin beamtests,
abletoreadtheabovefront-endelectronicsanddeliverthedatadirectlytotheEthernetnetwork. Section
5presentsthebeamtestsdonewiththefullreadoutchain,withbothplasticandliquidscintillatoroptions.
In section 6 we propose a target tracker architecture where the photodetectors are at the corners. In
sections 7 and 8 we discuss the implications of the electronics for the triggering and event-building
data-acquisitionandwespecifyfurtherourproposalforafullyEthernetbasedDACQ.Therstrealistic
nite-statemachinesimulationsofafullyspeciedDACQarchitecturearealsoshown. Inthenalsection
werecapitulateourndingsandpropositionsonallitemstouchedupon.
2 HPD characteristics
TheHPD'sarehybriddevicesusingaphotocathodetoconvertphotonsintophoto-electrons,accelerating
thelatterwithatypicalhighvoltagerangingfrom 8to 10kV.The8-10KeVphotoelectronsimpinging
onasiliconsensorconverttheir energytocharge. Theenergyneededto createan electron-holepairin
siliconisE
e h
=3:62eV.ThegainoftheHPDisgivenbyG(e ) = (U U
0
)(V)=E
e h
(eV),where
U istheappliedhighvoltageand U
0
representslossesinthesilicondead-layer;thenon-sensitiveregion.
ThetypicalgainoftheHPDis3000correspondingtoachargeof0.5fC.
Detailedstudieshavebeenperformedin Lyonusing a61 pixelsHPD model, produced byDEP[2],
thecharacteristicsofwhicharedisplayedin Table1. Fig.1showsthedesignofthe61pixelsHPD.Each
pixelhasitsownfeed-throughand connectingpin
1 .
TheexternaldimensionsoftheHPDare38mm13mm. Theusefuldiameterofinputwindowis18
mm. Thehighvoltageandgroundconnectionsuse2wiresontheside. TheHPDtubeissurroundedwith
arubbercoatingforisolation. Theinputwindowlies1mmunderthesurfaceof thiscoating. Thisis
usefulfortheconnectionofthebercookiewiththeinputwindowwithoutimposingstrongconstraints
onthetubeitself.
Thephoto-cathodeisdepositedonaberopticsentrancewindow. Thisgivesverygoodperformance
fortheHPDintermsofopticalcross-talk(seeSect. 2.5). Thephoto-cathodeisofmulti-alkaliS20type.
ItsquantumeÆciency asafunction ofthewavelengthisdetailedin Sect. 2.1andFig. 4.
InthisproximityfocusedHPDtheaccelerationdistancebetweenphoto-cathodeandthesiliconplane
isafewmillimeters. ThismakestheHPDimmunetomagneticelds. Thesiliconsensorisdividedin 61
hexagonalpixels. Eachpixelis 2mm sideto side,with anactiveareaof3.5 mm
2
. A depletionvoltage
of60Visrequiredtofullydepletethesilicondiode. Thegapbetween2adjacentpixelsis50mandthe
electricalcross-talkislessthan1%. Theoutputcapacitanceofthepixelsisaround4pF.
Thefront-endelectronicsoftheHPDshouldhavehighgainandlownoise. Thelowoutputcapacitance
allowstheuseof standardcommercialchipsoriginally designedforthereadoutof silicondetectors(see
Sect. 3for a complete description of the front-end electronics). We useVA-TA type electronic chips,
directlybondedonafront-endboard.
ManytestshavebeencompletedtocharacterizetheHPDintermsof: resolution,uniformity,linearity,
cross-talk, dark current with LED pulsed light and MIPS passing through the proposed scintillator
detectors(plasticandliquid);wepresentthem inthefollowing.
Table1: HPD characteristics.
Opticalcharacteristics
photocathodetype S20
quantumeÆciency (at480nm) 16.7%
quantumeÆciency (at520nm) 13.0%
usefuldiameter 18mm
inputwindow beroptics
operatingvoltage(max) -12kV
Electricalcharacteristics
totalactivearea 225mm
2
numberofpixels 61
atto atdistance(hexagon) 2mm
diodeactiveareaperpixel 3.5mm
2
gap betweenpixels 50m
depletiondepth 300m
depletionvoltage 60V
outputcapacitance 4pF
deadlayerthickness(junction side) 2m equivalentSi
board (9.5cm 12.5cm) andkeptin ametallic black box. Thelightcan be distributed to the HPD
either throughasingle1mm diameter clearberin contact withits entrancewindowor througha 61
pixels cookie matching the pattern of the HPD pixels (Fig. 2). This permitted us measurements of
thecross-talkwith aber-to-photo-detectorconnectionveryclosetothenal design. Inparticularone
hastoa)minimizethedistance betweentheentrancewindowplaneandthecookie andb)optimizethe
matching betweenthe bersand the HPDpixels. The second requirementis obtainedbyrotating the
cookieandadjustingthelightlevelontheilluminatedpixelandtheneighbours.
Figure 2: Left: HPDtest setupinLyon. TheHPDisinsidethemetallicbox. Theclear berbundleisvisible
ontopofthisbox. Also shownistheVA-DACQacquisitionsystemfromIDEASfortheVA-TAreadout. Right:
details of the HPD cookie collectingbers at theHPD entrancewindow. The hexagonalpattern matches the
HPDpixeldistribution.
We used LED ref. RS 235-9922 for the light source. This is the same with the one used by the
MINOS collaborationfortheir calibrationsystem[3]). The emissionof theLED's (470 nm) is closeto
the maximumemission peak of theWLS bers. The LED wasdrivenwith short pulses (around 50ns
width)ofvariableamplitudes anddierentrepetitionrates.
Thehigh voltage supplymust haveextremelylowripple. Weadopt theMatsusada[4] HV-15N-HP
with 3 mV ripple for 15 kV. The bias voltage is provided by a oating power supply. These power
supplies,aswellasthefront-endelectronicsdigitalandanalogsupplies,arecontrolledwithlowvoltages.
2.1 Quantum eÆciency
QuantumeÆciencyisobviouslyakeyissuefortheOPERATT.Thegreenextended S20photo-cathode
providesahighquantumeÆciency(above15%)atthemaximumemissionpeakoftheWLSbers. Fig.
3 shows the absorption and emission spectra of the Kuraray Y11 bers, dened as the baselineWLS
bersforOPERA.FurthertheseWLShavemaximalattenuationlength(7m)atthegreen. SeeSect.
5.2.1fordetailedresults.
Boththesefeaturesoersome exibilityintheTTdesign. Agoodphotonbudgetandlowattenuation
oerthepossibilityofputtingthephotodetectorsat thecorners(seeSect. 6).
TheDEPspecicationsgaveaquantumeÆciencyattheemissionpeakoftheWLSber(480nm)of
16.5%. An independentmeasurementhasbeenperformed by J.Seguinot at CERN [5]whohas kindly
agreedto comparethequantumeÆcienciesoftheLyonHPDfromDEPandtheLyon64pixelMaPMT
from Hamamatsu. Fig. 4showstheresultofthemeasurement. At480nmthequantum eÆcienciesare
Y11 absorption and emission spectra
0
20
40
60
80
100
300 325 350 375 400 425 450 475 500 525 550 575 600 625 650 675 700
Emission
Absorption
λ (nm)
Spectrum (a.u.)
Figure3: KurarayY11WLSbersabsorptionandemissionspectra.
HPD and MaPMT quantum efficiencies
0
0.05
0.1
0.15
0.2
0.25
0.3
200
250
300
350
400
450
500
550
600
650
700
750
800
λ (nm)
Quantum Efficiency
Figure 4: QuantumeÆciency of theS20 photo-cathode measuredby DEP (black dots) and by J.Seguinot at
CERN(redsquares). Also shownis the CERNmeasurement of theHamamatsu64pixels MaPMT QE (green
triangles).
OnewouldhavehopedtohaveabetterratiooftheQEforthe"multi-alkali"(HPD)overthe"bi-alkai"
one(MaPMT).Thefact that theyarecomparablecouldbeattributed tothediÆcultyofdeposition on
theberwindow,and thefurther lossesthatthethelatterimposesontheincomingphotons. DEPcan
guaranteephotocathodeeÆcienciesof thepresentrange,but not50%higher.
2.2 Resolution
HPD'sareknowntohaveexcellentphotonresolution. Thisistheconsequenceofthedissipativeprocess
TheresolutionisillustratedinFig.5obtainedby ashingagivenpixeloftheHPD(usedfortriggering)
with an average of 13 photo-electronsand looking at thesignalon the6 direct neighbours. Light was
injectedthroughasingleclearberatadistanceofroughly1mmfromtheHPDentrancewindow. The
typicalemissionconeangleoftheberusedis35
o .
Fig.5showstheresultonthe6neighbouringpixelsandthedetailofoneofthem. Thephoto-electrons
areclearlyvisibleandclearlydistinctfromthepedestal(rstpeakaround200mV).Atleast6peaksare
visible. Themeannumberofphoto-electronscanbeevaluatedeitherwiththemeanvalueofthepedestal
subtractedspectrumorwithaPoissonttothedistribution. Bothmethodsgiveforthistestanaverage
numberof 3photo-electronsonallthepixels.
Itisclearfromgure5thattheintrinsicresolutionoftheHPDgivesanautomaticcalibrationofthe
readoutchain. Thedierencebetweentwoconsecutivephotoelectronpeaksgivesthegain(typically 60
mV/p.ewiththefront-endelectronicsusedinthesetests). Duetodissipativesourceofmultiplicationno
gainvariationsareexpectedat thephotodetectorlevel.
0
250
500
750
1000
1250
1500
1750
2000
100
200
300
400
500
600
700
800
ID
Entries
Mean
RMS
13
60066
415.7
117.9
VA output (mV)
Number of events
Figure 5: Left: lightspectraonthesixdirectneighboursofa ashedpixel. Right: detailedspectrum ofoneof
thepixels.
TheHPD resolution ofthe singlephoto-electron has alsobeen achieved in auto-triggermode. The
thresholdoftheelectronicswassetbetweenthepedestalandtherstphoto-electronpeakonthetriggering
part. We adjust this value on each pixel to compensate for pedestal spreads. Fig. 6 shows the dark
current spectrum of the photo-cathode of one channel. The signal-to-noise ratio (around6) gives an
eÆcientseparationbetweenpedestal and rstphoto-electron peak. Since onetriggersona partof the
signalshaped withafast amplier thenoiseishigher atthe triggerpartthan at thecharge-measuring
partandthegoodsignaltonoiseratioofg6doesnotmeanthatwetriggerwitha100%eÆciency. The
auto-triggereÆciency, iscurrentlyunderstudy(see3).
2.3 Linearity
We studied the evolution of the HPD gain as a function of the high voltage applied. This test was
performedbyrecordingthedarkcurrentspectrumforvarioushighvoltages. Theminimumhighvoltage
valuetoseparatetherstphoto-electronfromthepedestalis5kV. Themaximumhighvoltageapplied
was10kV.Thepositionsofthetwopeaksaregivenbythemeanvaluesoftwogaussians. Thetheoretical
gainslopeis 1000=3:62=276:2 e perkV. Theresultof themeasurement isdisplayedonFig.7with a
Dark current spectrum (HV=9kV)
0
250
500
750
1000
1250
1500
1750
2000
100
150
200
250
300
350
400
450
500
ID
Entries
Mean
RMS
13
17484
274.5
28.06
170.5 / 57
P1
1820.
P2
254.2
P3
9.230
P4
968.9
P5
303.8
P6
11.11
VA output (mV)
Number of events
Figure 6: Darkcurrentspectrumobtainedin auto-triggerablemode
Slope = 268 e
-
/kV
Figure 7: Left: Positions of the pedestalsand therst photo-electron peak as a functionof the highvoltage
appliedtotheHPD.Right: Evaluationofthegainforeachhighvoltageapplied.
2.4 Uniformity
Thepixel-to-pixelgainuniformityis aclearadvantageof theHPD.Theuniformityisimportantforthe
front-endelectronicsdesignthathasnotto takeintoaccountthegainspreadcompensations.
Wemeasuredtheuniformitywiththedarkcurrentspectrumforeverypixelasexplainedintheprevious
subsection. Weoperatedin auto-triggerable modewith athresholdadjusted between thepedestaland
therstphoto-electronpeak. Thehighvoltageappliedwas10kV.Thegainobtainedwasthencorrected
withthevariationsin gainofthefront-endelectronicsdeterminedindependentlyinachargecalibration
mode. Theresultis shown in Fig.8. Theleft plotdisplaysa2-D viewofthe 61HPDpixels with their
gain(inelectrons). Therightplotgivesthedistribution ofthesinglephoto-electronchargeQ
pe
for the
61pixels. Theresultinggainnon-uniformity,denedas
Q pe
=hQ pe
charge
0
2
4
6
8
10
12
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
13.67 / 11
Constant
7.277
Mean
.4730
Sigma
.2738E-01
charge
0
2
4
6
8
10
12
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
13.67 / 11
Constant
7.277
Mean
.4730
Sigma
.2738E-01
charge (fc)
number of channels
Figure 8: Left: projectionofthe gain(inelectrons)forthe 61pixelsofthe HPD.Thegain spreadrangesfrom
2500to3500electrons. Right: singlephoto-electronchargedistribution.
2.5 Cross talk
Weevaluatedtheopticalcross-talkoftheHPDbyilluminatingapixelwithaclearberincontactwith
theentrancewindow. Thelightinjectedhadanaverageof6photo-electrons. Noopticalgreasewasused.
Thecross-talkisgivenbythefraction ofsignalreceivedbytheneighbouringpixels.
The Fig. 9shows the 3-D viewof the signalreceived by the 61 HPD pixels, normalizedto the signal
of the ashedpixel. Themaximum signalcorrelatedwith thelightinjectedis recordedon the6direct
neighboursofthe ashedpixelandisnotlargerthan2%perpixel. Thisintrinsiclowcross-talkisobtained
thankstotheberoptic entrancewindowwhichgivesabetterfocalisationoftheincominglight.
Normalized mean signal per pixel
Cross talk
mm
mm
Signal in triggered pixel (mV)
Signal in neighbour pixel (mV)
Figure9: Left: 3-Dviewofthesignal receivedbytheHPD61pixels,normalizedtothemaximumsignal. The
maximumcross-talksignalislessthan2%. Right: distributionofthelightreceivedbyapixelw.r.t. thedistance
Thecorollaryofthesensitivityatlargewavelengthsistheincreaseofthedarkcurrent(thermo-emission
of the photo-cathode). This darkcurrentmust be kept at a reasonablelevel in order to minimize the
dead time in an auto-triggerable acquisition operating with athreshold of afraction of photo-electron
(seeSec.7fordetailed results).
Two dierent measurements have been performed to cross-check the dark current rate with the
onegiven by DEPspecications. The rst oneis a direct measurementof the dark current spectrum
in auto-trigger mode with a threshold below the rst photo-electron. The limitation of such a
mea-surement is the uncertainty on the trigger eÆciency at any given threshold. A second measurement
has therefore been performed with an external trigger by looking at random dark current counts
occurring in correlation with the external trigger. Indeed the number of such events is given by :
N =f DC f external T integration ,wheref DC
isthedarkcurrentrate,f
external
theexternaltriggerrate
andT
integration
isthetimeofsensitivityoftheelectronics(typicallyfews). TheparameterT
integration
and the cut applied to select events corresponding to single photo-electrons have been cross-checked
withaLEDspectrum(pulsedlightatafrequencyequaltof
external
)). Fig.10showsthedierentspectra
obtainedwithsuchamethod.
500 s of random trigger at 10 kHz
ch 14, pix 32
Slow shaping VA (mV)
N/4 mV
ch 14, pix 32
ch 14, pix 32
Figure 10: Light and dark current spectra obtained with an external trigger. The light spectra is used for
calibration. Thedarkcurrenteventsareintherstphoto-electronpeak.
Bothmeasurementsare consistentwithadarkcurrentratearound25 kHzcm
2
or1kHz/pixel. It
shouldbeemphasizedthattheHPDwasinsideitsclosedmetallicboxandthetemperaturewastypically
T ambiant
+5
o
. Theresultsobtainedin LyonareconsistentwithDEPspecications(14kHzcm
2
)given
theuncertaintiesonthetemperatureconditions(thedarkcurrentmaydoubleevery5degrees).
Thishighvalueofdarkcurrentrateisclearlyunacceptable. DEPagreedtotestaanewtypeofS20
photo-cathode,the so-called"hot"S20,with samefeatures around 500nmbut lowersensitivity in the
red(seeg11). Thespecicationsforthedarkcurrentarebelow4kHzcm
2
. Thisnewphoto-cathode
isavailableonthe61 pixelsversionand isalsothe photocathodeusedbythenewmultipixel HPD: the
163pixels HPD(see Sec. 6 fordetails). Given thelower size of the pixels in this new HPD, the dark
currentrateshoulddecreaseto anacceptableratebelow65Hzperpixel. Anew61pixelHPDwith"hot
S20 photocathodes quantum efficiencies
0
5
10
15
20
25
30
200 250 300 350 400 450 500 550 600 650 700 750 800 850 900
S20
Hot S20
S20 UV
λ (nm)
QE (
%
)
Figure 11: Left: quantumeÆciency oftheS20, "hot"S20 andS20 UVphoto-cathodes asmeasured byDEP.
TheS20 UVphoto-cathodemaybedepositedonaquartzinputwindowonly.
2.7 Advantages and limitations of the HPD for OPERA
TheHPDadvantageshavebeenpresentedindetail:
Excellentresolution(S/Babove6)permittingautocalibrationofgain
Excellentuniformity(2%)andlinearity(1%)
Verylowcross-talk(2%)
Immunityto magneticelds. ThoughthismightbelessrelevantforOPERA
Thetwo"weak"pointsoftheHPDare:
thehighrateofdarkcurrent(1kHz/pixel),adeciencythatwillbehopefullycorrectedbythenew
photocathodethe"hotS20"(65Hz).
the low gain, imposing sophisticated electronics. This we consider a less severe problem. We
believethatthenewgenerationoflownoiseelectronicsissuÆcientlywellunderstoodpermittingus
totacklethisproblemwithsuÆcientcondence. Infactwehaveshownintheprevioussubsections
resultsonthesinglephotoelectrondetection. Furthertheexperimentallevelsofnoiseachieved(see
3.1 The VA-TA principle
Thefront-endelectronicsbelongtotheVA-TAfamily,rstdevelopedatCERNunderthenameVIKING
[6] and produced byIDEAS [7]. We brie y recallits principle: the VA-TA's are multi channel chips
2 ,
with charge-sensitivepreampliers followed by afast and a slow shaper. The slow shaper (VA) has a
peakingtime of 1-2microsecondsand thefast shaper(TA) hasarise-time of75 nanoseconds. Each
channel inthetriggerchipTAisfollowedbyalevel-sensitivediscriminatorwithadjustablethreshold,a
serialshift-registerto selectthechannelsallowedtotriggerandamonostable. IntheVATAcgversiona
dierentthreshold canbeset foreachchannel. Theoutputsare ORedtogetherand canbesentto the
VApart. TheOR thenservesas asampleand hold signalforthechargeaccumulatedat theVA part.
Thedelaybetweentriggerandhold matchesthepeakingtimeof theVA (2s). ThesameORsignal
caninitiate the readoutsequence. IntheVA part,theoutputs of thepreamplier enter amultiplexer.
Itsswitchesarecontrolledbyabit-registerrunninginparallel. Theoutputofthemultiplexergoestothe
outputofthechipviaabuer. Onlyoneofthechannelsisseenontheoutputateveryreadoutstep. The
speedofreadoutisthenlimitedbythemultiplexerclockwhichmakesthebitintheregistertogglefrom
onechannelto thenext. Thespecicationsforthemaximumacceptable clockfrequencygive10MHz,
thoughweencounteredproblemswhentryingtoreadabove5MHz. Thedata canthenbeclockedinto
anADC.
Thechipcanalsobeoperatedintestmodeviaanothermultiplexer/bit-registerontheinput,injecting
charges in the preamplier inputs. Further, it is also possible to use an external trigger for the VA
sample/holdand thestart of thereadout sequence. We willreport below ontests where wehaveused
bothaninternalandanexternaltrigger.
TheworkingprincipleandthetimingsequenceofaVA-TAcombinationaredisplayedinFig.12.
O
U
T
P
U
T
M
U
L
T
I
P
L
E
X
O
R
T
E
S
T
P
U
L
S
E
M
U
X
00
00
11
11
pad
00
00
11
11
H1(s)
H2(s)
+
-H1(s)
H2(s)
+
-TA
VA
Preamp
(charge
integrator)
Slow
shaper
Fast
shaper
Discri-minator
Fixed
with
monostable
vss
vss
vdd
Test pulse
Hold
Analogue
output
ORed
trigger
output
S&H buffer
Input
Time
Signal
Slow shaper
Fast shaper
signal
Preamplifier
Physical event
Undetected trigger
Threshold
4
3
N-1
N
2
Clock
1
Trigger output
Hold
Figure12: WorkingprincipleandtimingsequenceoftheVA-TA.
GiventhelowgainoftheHPD(3000e perphoto-electron)theVA-TAshould havehigh gainand
lownoiseperformance. Thesecharacteristicshavebeeninvestigatedeither in testmode(withinjection
ofchargeviaa1.8pFcapacitance)ordirectlywiththeHPDconnected. Wehavealsocharacterizedthe
chipsintermofdynamicrange,peakingtimesandcommonmodenoise.
2
The VA-TA concept can beapplied both to HPD and MaPMT. Onehas to adjust the polarity of the
inputsignalandthedynamic rangeofthechips. FortheHPDreadout, weusetheso-calledVA32cand
VA32rich. The rstone has agood signal/noiseratiobut averylimited dynamic range(linear on the
11.5fC range, corresponding to a maximumlight yield of 46 photo-electrons). The second one has
alarger dynamic range (linear from -65 fC to +65 fC, correspondingto 260 photo-electrons). For the
MaPMTreadoutwecanusethetheVA32hdr11chip(hdrstandingforhigh dynamicrange). Although
the general features of this chip are comparable to the one of the former chips, its dynamic range is
considerably largerandmatchesthe characteristicsof aPMsignal(from -25pC to+25 pC).All these
chipswerebondeddirectlywithaTA32cg chipfortriggering.
Thecomparativedatafor theVA32c,VA32rich, VA32hdr11andTA32cgare giveninTab.2. Dueto
lackoftimewewereabletotestfullyonlytheVATAcgchip.
Table2: VA-TAperformance.
Chip VA32c VA32rich VA32hdr11 TA32cg
Gain(mV/fC) 150 35 0.1 -16(Gainstage=10)
Noise(mV) 10 1 0.25 2.5
Peakingtime(s) 2 2 0.8 0.075
Dynamicrange(fC) 11 65 25000
3.3 Performances of VATAcg
3.3.1 Gain
Thegainismeasuredincalibration mode. SymmetricmeasurementsareperformedfortheVAand TA
chips. For the VA chip, the pulse-heightis recorded for dierent valuesof injected charge. This gives
the value of the pedestals and the gain. For the TA, a scan is performed over the threshold voltage
rangeforagiven calibrationcharge. Thethresholdis foundwhenthe 50%triggereÆciencyis reached.
Therepetitionoftheprocedurefordierentcalibrationchargesallowsthedeterminationthegainofthe
chip. The symmetricprocedure mayalso beused for cross-check(scan of thechargerange foragiven
voltagethreshold). Bothprocedures leadtosimilarresults. TheTA32cgchiphasagainstagethat can
beadjustedexternally. Thisstageallowstoselectvariablegainintheapproximativeratios1:3:6:10. The
distributionofthegainsfortwoVA32candtwoTAcgchipsaredisplayedinFig.13. Themeanvaluefor
the VA(TA) is around120(-16) mV/fC and the gainspread/chipis around2%. Thepedestalsremain
stablewhentheHPDisconnectedandtheHV isapplied.
3.3.2 Noise
TheVAnoiseiscomputedfromthetofthepedestaldistribution. ThedistributionsfortwoVA32cchips,
withoutcommon-modenoisesubtraction,aregiveninFig.14. Theyarearound10mVcorrespondingto
500e ENC.Theintrinsicnoiseofthechip(obtainedbeforeconnectingtheHPD)isoftheorderof200
e ENC.TheTAnoiseiscomputedfrom thetriggereÆciencycurve. Atypicaldistribution isshownin
thesamegure. Itisaround2.5mVcorrespondingto900e ENC.Fromstandardtheoreticalcurvesof
noiseversussamplingtime oneknowsthat afactorof2reductionin noiseforboththeVAand theTA
is possible(200 and 500e ENC respectively for theVA and theTA); provided an optimized PCBis
designed. FortheVA, theperformancesastheyarenowaresuÆcient,butareductioninnoiseoffactor
2wouldcertainlybemorethanwelcomeintheTApart.
3.3.3 Trigger
TheTAgivesatriggerthatiscommonforallthechip. ThethresholdisalsosetforalltheTAchannels.
Thechannel-to-channelspreadbecomesalimitationwhen weare dealingwithverylowthresholds. To
distribution of gain
0
1
2
3
4
5
6
7
8
9
10
80
90
100
110
120
130
140
150
160
24.83 / 13
P1
3.931
P2
116.2
P3
2.099
P4
4.288
P5
123.7
P6
2.622
gain (mV/fC)
number of channels
100
150
200
250
300
350
400
0
10
20
30
40
50
60
channel
pedestal (mV)
Figure 13: Distributionof the VA32c gain (left),TA32cg gain (center)and VA32cpedestals (right). Thelast
plotshowsthestabilityoftheVApedestalswhenweconnecttheHPDandapplythehighvoltage.
distribution of noise
0
1
2
3
4
5
6
7
0
2
4
6
8
10
12
14
16
18
20
9.751 / 15
P1
6.581
P2
9.006
P3
-.3608
P4
3.092
P5
12.04
P6
.8084
noise (mV)
number of channels
Figure14: DistributionoftheVA32c(left)andoftheTA32cg(left)noise. Thevariationofthenoisew.r.t. the
peakingtimeisalsoshown(right).
of the thresholdsaround the commonvalue. Thethreshold spreadis then reduced bya factor 3after
optimizationofthebit-patternin theDACmask(Fig.15)
3.3.4 Dynamicrange
Thelinearityofthefrontendelectronicsoverthefullrequireddynamicrange(1 200p:e:corresponding
to0:5 100fC)canbecheckedinthecalibrationmode.
ThesignaloftheVAasafunction of theinputchargedeterminesthedynamic range. Thedataare
wellttedbyapolynomialoforderthreeVA
output = a 0 +a 1 Q+a 2 Q 2 +a 3 Q 3
. Fig.16showsthedataof
thesignalandthetasafunctionoftheinputcharge. TheVA32Chasalineardynamicrangefrom0fC
to12:5fC. Forhighervalues,thedierencebetweenalinearandapolynomial(oforderthree)response
becomes important, but acceptable and can be corrected up to 20 fC. With minor modications the
chipcanacceptonlyonepolarityandthereforedoubleitsrangeupto25-40fCcorrespondingto80p.e.
Butthis isnotthedirection we havetaken, sinceweintend to testtheVA-rich chipwhich hasamuch
Mean = -45.91 mV
σ = 2.491 mV
Mean = -46.86 mV
σ = 0.802 mV
Figure15: ReductionofthethresholdspreadafterDACoptimizationintheTAchip.
Non linearity (‰)
Duringthebeamtestsof Mayand June2001wecompleted thersttests ofafullreadout chain, from
thescintillatortotheworkstation,inwhatweproposetobethenalconguration. Theresultsofthese
tests will be presented in the next section. In this section wepresent the acquisition card, under the
preliminary nameORCA(OperaReadoutCArd) which includesa12bit,25 MHzADC, thesequencing
element(aFPGAofthetypeAlteraAPEX20K200)andanEthernetchipportingthedatadirectlytoa
PC.TheEthernetcontroller(currentlya10Mbits/sHPchiptobesoonreplacedbya100Mbit/sLINUX
chip)allowsnotonlytoaccessandcontrolthefront-endelectronicsthroughhttpperformingslowcontrol
functions but alsosend thedatathroughTCP/IP[8] totheworkstationwith highthroughput (forthe
timebeing: 1Mbit/s). Thedesign,development,productionandtestingofORCAhasbeendonebyour
laboratory.
AblockdiagramofORCAand itsinteractionwiththeVATA front-endisshowningure 17
From Photodetector
0
63
Or trigger
Readout logic signals
PPS
Ethernet
Analog
Front End
ADC
Stamping
Time
Logic
Control
FIFO
Ethernet
Capable
Device
Calibration
Monitoring
...
Optional Hard Reset
Figure17: SchematicsofanEthernetCapableFront-endModule
Thiscardallowsadirectcontroloftheacquisitionwithoutusinganacquisitionbus. Itiscosteective,
giventhelowpriceoftheEthernetcontrollerchips,easyto controlfromaPCunder Windowsor Linux
andwellsuitedforacompleteintegrationclosetothefront-endelectronicssinceonlyoneEthernetcable
is needed to access the acquisition. It is also a perfectly exible candidate for testing the scintillator
modules atdierentlabsduring theproductionphase,andduringinstallationat GranSasso, sinceitis
perfectlyautonomous: anyportablePCcanbeconnectedtotheEthernetportandtakedata.
Aphotoofthecardcanbeseenin gure 18.
ORCA, in moredetail,is anEthernetCapableFront-endModule usingaEthernetdevicefrom Agilent
Laboratories[9](WebplugorBFOOT11501). ThisdeviceisacustomASICbasedona68000
micropro-cessorandtheVxWorksrealtimekernel. ItcontainsanembeddedEthernetcontroller,includingaWeb
serverandexpectstoreceivethefront-enddatathrougha5MbpslinkobeyingtheIEEE1451.2standard
[10]. TheIEEE1451.2standardprovidestheabilitytoproducenetwork-capable"smartsensors". It
par-titionsasmartdevice intoaNetworkCapable ApplicationProcessor(NCAP, inourcasethe BFOOT)
andaSmartTransducerInterfaceModule(STIM,inourcasetheADCwiththeFPGAsequencer). The
STIM architectureincludes twomain blocs. Therst blocimplementsthe minimal mandatorypartof
the IEEE 1451.2 standard. The second main bloc containsthe transducer channels and the front-end
sequencer. Thisblocprovidesallthefeaturesneededtocontroltheanalogfront-endchip(VA-TA32cg).
TheNCAPprocessorinterfacestheSTIMwiththeNetwork andallowstocontrolittransparently.
Thepresentprototypeincludesatimestampingfunction,realizedbylatchinganexternalsignaland
givingitatime labelcomingfrom alocalclock. A dedicatedinpu fromanexternally distributed clock
synchronizedonaGPSreceiverpermitsanabsolutetimestamping.
Anodecanbeaccessed inavarietyofways.
Theprimarymethod isthroughthewebinterfacebytypingaUniversalResourceLocator(URL).
Fig. 19 shows an example of JAVA appletwhich can read continuously an ADC channel in an
oscilloscopemode.
Figure19: WebOscilloscope
The node can also be accessed by http commands. E.g the command: Http
://ly-otmp9/bin/1451dot2/read ?startChan=2&stopChan=2&stim=1 reads achannel with 2
bytes ofdata.
TheBFOOTalsoprovidesadatastreamingapplicationwhichallowstokeepanetworkconnection
openedandsendthedatacontinuouslyonEthernet. Thedatastreamingmodeisaspecic
applica-tionbasedontwothreads,which sharethesamesetofringbuers. Wehaveachievedthroughput
close to 1 Mbit/s with this function. Figure 20 shows an example of data taken through the
network.
We validated the concept using the embedded Ethernet controller BFOOT 11501. Unfortunately,
AgilenthasdecidedtodiscontinuethischipduetoinsuÆcientdemand. Weluckilyfounditssuccessor,a
lowpowerprocessor(less than500mW),implementingfullLINUX capabilities, includingmultitasking
and equipped with a 100Mbit/s port. It is called ETRAX [13] and has had a very wide distribution
(over1millionsoldalready). Choosing aprocessorwithLINUX, isaguaranteethatwewillbeimmune
toindustrial accidentsofthetypeexperiencedalreadywithAgilent.
4.2 May-June beam tests: test of the full readout chain
wasachievedviaa50pinsERNIconnectoranda atcable(thiswasimposedbythefront-endelectronics
board). TheHPD,pluggedonthefront-endboard,wasinsideasmallmetallicboxshieldedagainstlight
leakages. Allthepowersupplies(high voltageandHPDbiasvoltage)werexedonthebacksideof the
module. The module was thendriven by lowvoltage signals(max+24V) controlled from the control
roomandthedataacquisition wasperformedthroughasingleethernetconnection. Thisgaveusahigh
mobilityand exibilityofinstallationandrunning.
Figure 21: Picture oftheHPD/VA-TA/EthernetDAQmodule usedduring theMay and Junetestsbeamsat
CERN.The HPD withitsfront-endelectronicsis located inthe metallicbox. The DACQcard is sandwiched
betweenthetwometallicplanes. Powersuppliesareattachedontheback-plane.
We tested the control of the front-end electronics both in external and in auto-triggerable modes.
We checkedthe data transferup to 1Mbits/s (corresponding to 150Hz/channel eventrate for the 64
channel VATAcg). Wealsocheckedthetimestampingandlabelingoftheevents. Thetriggers(external
and auto-trigger)weretimestamped thanksto alocal50 MHzlocalclock (20nsaccuracy) toallowfor
o-linereconstruction.
Thedata transferred were the ADC value(s) for 1 to 64 channels, the triggertimestamps and also
two dierent types of counters giving the number of auto-triggers between two external triggers and
thenumberof auto-triggersin agiventime windowafter thelast externaltrigger. These twocounters
are used to study noiseand triggereÆciencies. All thefunctionalities of the readout chain havebeen
successfullytested. Datapresentedin thenextsectionwereobtainedwiththisreadoutsystem.
0
1
0
0
1
1
HPD cookie
Logic
Stamping
Time
Logic
Control
Calibration
Monitoring
...
ADC
Ethernet
Capable
Device
FIFO
H1(s)
H2(s)
+
-H1(s)
H2(s)
+
-Preamp
(charge
integrator)
Slow
shaper
Fast
shaper
Discri-minator
Fixed
with
monostable
vss
vss
vdd
Test pulse
Hold
S&H buffer
Input
pad
T
E
S
T
P
U
L
S
E
M
U
X
O
U
T
P
U
T
M
U
L
T
I
P
L
E
X
O
R
VA
TA
signals
ORed
trigger
Analog
out
HPD
61 pixels
2 x 32 channels
VA-TA
DAQ
Ethernet
(48 connected)
Scintillator
(3 x 16 strips)
optical
connector
Figure22: Generalarchitectureofthescintillatortelescopereadoutchain. Themajorelementsofthechainare
thedetectoritself,the HPD,theVA-TAfront-endchipsandEthernetDACQsystem. Thefull chainwastested
Weperformed testsof scintillator readoutwith anHPD, usingbothplastic and liquidscintillatorwith
radioactivesources,cosmic raysandparticlebeams.
5.1 Tests of the HPD with the plastic scintillator
5.1.1 Asmall telescope prototype
A small plasticscintillator prototypewasbuiltin Lyon. The prototypehas3planesof 16plastic
scin-tillatorbars. Thescintillatorcame fromPol.Hi.Tech(same batchastheoneusedto buildthefullscale
prototypeinStrasbourg[11]). Thebarswere41cmlong,2.6cmwideand1cmthick. Eachbarisread
by WLS bers. Two planeswere equipped with Kuraray Y11 bers(1.2 mm diameter) and the third
onehas Bicron BC91Abers (1 mm diameter). The WLS bers end to an optical connector at both
ends. Theyaregluedtothisconnectorwithcarbondopedglue. Theendsofthebersarethenpolished
by hand. The minimal and maximal ber lengthsare respectively 70 cm and 100 cm. The bersare
gluedto thescintillatorin agroove(1.6mmthickand1.6mm deep)made withasaw. Theglueis the
standardBicronBC600. Thebarsarethen paintedwithare ectivepaint. Weused theBicron BC620
and applied 4layers with abrush. The strips are then collected, glued together in a 70 cm 41 cm
U-shapealuminiumprole(the aluminiumthicknessis500m). Theopticalconnectors arealso glued
onthisprole. TheplaneisthenclosedwithanotherU-shapealuminiumprolegluedtotherstprole
andtothestrips(Fig.23).
Figure 23: ViewofascintillatorplanewithstripsgluedonanaluminiumproleandtheWLSbersendedby
thersthalf ofanopticalconnector.
Thethree planescanbe alignedthanksto transversebarsxed withscrews. Thedistance between
5.1.2 Connectortransmissionand crosstalk
InthetestsofMayandJune2001atCERNthelightwasreadononesidebyasingleHPDwith48pixels
connected. Thetransmission wasensured by abunch of 35cm longclear bers(Bicron BC98MC,1.2
mmdiameter) endedonthe telescopeside byanopticalconnector, matching theoneof thescintillator
plane,andonthephoto-detectorsidebyahexagonalcookiematchingtheHPDpixelpattern.
Thestrip-to-clear-berconnectorshavebeendesigned andbuiltin Lyon. Theyconsist of twoPVC
pieces, 68 mm large and 10 mm wide with 16 holes for the bers, two 3 mm diameter holes for the
alignmentpinsand two3mm holesforthe screws. Twoconsecutivebersare separatedby a2.5mm
gap. Fig.25givethedetailsoftheopticalconnector.
Therelative alignmentof thetwohalvesisgivenbytwometallic pins. The mechanicalaccuracyof
thebersholesissupposedtobeatthelevelof20m.Thesurfaceofthetwohalveshasbeenpolished
byhand. The overalltransmissionof these connectors hasbeen measuredto be above80 %. It canin
principlebeincreasedwithahardermaterial(PVCused toget somedeformationsifone pressthetwo
halvestoohard),opticalgreaseandalsobyadierenttreatmentofthesurfaces(withadiamond y-cut
machiningforinstance).
Onthecookiesidewemade"Mercedeslike"cookiestomergethe316berbunchestothesameHPD,
testing the possibility that the granularity of the detector does notmatch thegranularityof thestrip
module,orthestripconnector. Thenon-connectedpixelswerecoveredwithblackadhesivetominimize
possiblere ectionof light at thisend. We havealsoconstructedsimilar cookiesfor theMaPMT's(see
Fig.26. Theywillbeusedincomparativetests.
5.1.3 MIPresponse
TestshavebeendoneinCERN(PST7)inMayandJune2001. Therstcalibrationtestisthe
Figure26: PictureoftheHPD(left)andMaPMTcookies(right)connectedtotheplasticscintillatortelescope.
of 4scintillators(2 ofthem were1cm widengers). Werunwith10 GeVparticles(mostlypionswith
asmallamountof electrons). Theresponse ofthe stripsisdisplayedonFig.27. Twoestimations have
beenperformed, therst one using aPoisson t to thedistribution and the second onerelyingon the
meanvalueofthedistribution. Discrepanciesbetweenthetwoestimationsoccurforsomestrips(mainly
intheplanewith1mmbers)becauseoflargepedestalsonthecorrespondingchannelsthatgivealower
thanexpectedmeanvalue. Anotherlimitationin thesystematicscan ofthe MIPresponse wasthelow
statisticsfor thestrips at theedges (mainly strips15 and16). Theaverage numberof photo-electrons
obtainedontheplaneswith1.2mmWLSbersis8.4fortherstplane(withaRMSof0.8)and9.0for
thesecondone(withaRMSof1.3). Thecorrespondingnumberofphoto-electronsforthethirdplaneis
6.1(RMS1.2). Thislowervalueismainly duetothelowerlightyieldoftheBC91Aberscomparedto
theY11. Oneberwasbrokenat theedgeof theplane. The resultsobtainedhereareconsistentwith
plane 1
N
p.e.
per mip at d
HPD
=40 + 35 cm
Strip number
N
pe
plane 2
Strip number
N
pe
plane 3
N
p.e.
per mip at d
HPD
=40 + 35 cm
Strip number
N
pe
Figure27: MIPresponseof32stripsequippedwith1.2 mmWLSbers(leftplots)andwith1mmWLSbers
(rightplot). Theaverage numberofphoto-electronsisrespectively9and6forthetwoberdiameters. Fiber#
16onplane3isbroken.
5.1.4 Opticalcross-talk
Theoptical cross-talkwith theHPD cookie hasbeenstudied aswell. As mentionedbefore, one hasto
optimizethe position ofthe bersw.r.t. the position ofthe HPDpixels. This isdone byrotating the
cookiearoundthecentralbertomaximizethesignalononegivenpixeland tominimizethesignalon
the neighbours. This procedure givesasatisfactory cross-talk below 2%(which is the intrinsiclimitof
the HPD).Fig.28 showsthe cross-talkfor theeventsoccurring in pixels 13, 24and 42(corresponding
to thestrips# 8in each plane). Thefractionof theincidentlightis lessthan2%onthe neighbouring
pixelsofthese3pixels.
5.2 Tests of the HPD with the liquid scintillator
Tests have also been performed using liquid scintillator setups. The detectors have been produced at
CERN(byI.Kresloandcollaborators). Theyconsist ofpolycarbonatecells(1 1cm
2
cross-section)of
variable lengths (up to 7 m) lledwith BicronBC570 scintillating oil. Thelight is read by a1.2 mm
WLSber(Y11)whichislocatedapproximativelyatthecenterofthecell. Eachcellisclosedbya3
cmsiliconjoint(CAF50siliconglue). Wersttestedwith cosmicraysa12cellsdetector(4 planesof3
cells) 6m long,read ononesidebytheHPD,withoutanyre ectorontheother side. Inthisdetector,
thecellswere notclosedandthe liquidcamedownto theHPDcookie . Thenwetook dataonthefull
scaleliquidscintillatorprototypeat CERNduring theMayandJunebeamtestperiod. TheTiO
2 load
wasincreasedto20%betweenthetwoteststoimprovethelightyieldanddecreasetheopticalcross-talk
betweenadjacentcells.
5.2.1 Cosmicray setup. Measurementofattenuationlength
Thecosmic raysetupdescribed aboveis shownon Fig.29. Thedetector waswrappedwith aluminized
Mylar. The triggeringsystem consists of 3PM's in coincidenceand the trigger was send to the HPD
readoutsystem. The12berswereconnectedtoanHPDthroughaspeciallymadecookie.
Theaim of this test was to measure the attenuationlengthalong the ber. Data havebeen taken
at dierentpositions along the detector (from 0.6 m to 5.8 m) by moving the triggeringsystem. The
averagenumberofphoto-electronsat4m wasmeasuredtobe5. Fig.30showsthevariationofthelight
Cross talk, Plastic Module
Normalized average signal per HPD pixel mm
mm
Figure28: Displayofthe61HPDpixels. Selectedeventsoccurinpixels13,24and42. Thesignalsarenormalized
tothemaximumsignal.
Figure29: PictureoftheliquidscintillatorcosmicssetuptestedinLyonwithaHPDreadoutsystem
of cosmicraydata taking)thecurveisttedwith asingleexponential. Thecorrespondingattenuation
Attenuation length measurements w. Liquid Scintillator
Distance from HPD (m)
Mean nr of p.e.
Figure30: Variationofthelightyieldw.r.t. thedistancetotheHPD.Thettingcurveisasingleexponential.
5.2.2 Beamtests
The MIP response of the full scale liquid scintillator prototype was also investigated with the HPD
readoutsystem. Inthisexperiment,a6mhorizontalplaneof50cellswasreadbyoneHPDononeside
and one EBCCD on theother side. 2horizontal and 3 vertical planes were also read by the EBCCD
ononeside. Leadbricks(5.6 cmof Pb) were sandwiched betweenthe 3plane"doublets" andvariable
amountofPbwasputin frontoftherstplane(theoneconnectedto theHPD).Testswereperformed
atvariableenergies(2,4,6and 8GeV).
WepresentheretheresultsfortheMIPresponse. ThespectraforonecentralcellisgiveninFig.31.
The average number of photo-electrons 4.6 at 4m distance, is fully consistent with the one obtained
withthe cosmicraysetup atacomparabledistance. Themip responseand attenuationlengtharealso
consistentwitharadioactivesourcemeasurementnotshownhere.
ch44
Strip 23, plane 1
N
pe
N
Figure 31: MIP response spectrum at 4 mfrom the HPD ona central cell(# 23). The average numberof
Thebaselineoptionfortheelectronicdetectorin OPERAisascintillatortrackerwithplasticscintillator
strips,6.7 mlong,1cmthickand2.6cm wide,readoutwithWLS bersandthe Hamamatsu64pixels
H7546MaPMTplacedatbothendsofthebers[12]. Thegeometryofthetrackerisxedbythechoice
of these elements. The basic units are sub-modules of 64 strips readout by two MaPMT's. Four
sub-modules areassembled in situto constructawholescintillatorwall. Twoplaneswithstrips orthogonal
arecoupledtoformax y tracker. BasicnumbersforthebaselinetargettrackerarerecalledinTab.3.
Table3: Reminderofthenumberofelementsinthebaselinetargettracker.
# perplane 3super-modules
strips 256 36864
units 4 576
MaPMT's 8 1152
Inthisbaselinedesignthephoto-detectorsarecontainedin thesub-modules withthefront-end
elec-tronics. Themain advantagesofthis solution,is theavoidanceofopticalconnectors andthethe
modu-larityofthewholecomplex. Its maindrawbacksasweseethemare:
lowaccessibilityofthephoto-detectorsandthefront-endelectronics (inparticularforthevertical
stripsplanes);
fragilityofthemodules(withphoto-detectorsandelectronicsembedded)duringtransportationand
assembly;
closecouplingbetweentheproductionofthescintillatormodules andthephotodetectorand
elec-tronicsassembly. Weputbothitemsonthecriticalpath.
large longitudinal dimensions of the sub-modules. The end-manifold length, around 40 cm, is
imposedbythebersbendingradius,thephoto-detectorandcookielengths,theelectronicreadout
card andmechanicalsupport. Thisextralengthcouldinterfere withtheworkingsofthethebrick
manipulatorrobot. Areductionofthislengthcouldbewelcome;
large transversedimensions imposed by thephoto-detector. The end-manifold hasathicknessof
32 mm : 30 mm due to the MaPMT thickness and 1mm margin on each side. Probably larger
margins will be needed. The increased length increases quadratically the scanning load for the
eventsrequiringabrick-to-bricktrackmatching.
In this note we propose a possible alternative to the baseline option which foresees to locate the
photo-detectorsin the"corners"of thetarget trackerplanes. Thephoto-detectorsshould behoused in
metallicboxesattachedtothetransversebeamsofthedetector. Thissolutionallowstoovercomeallthe
drawbacksstatedbeforeatthecostof theuseof clearbersandoptical connectors(optionretained by
theMINOScollaboration,seeSec.6.3formoredetails).
Thetransverse dimensionscouldbereduced down to theintrinsic width of thescintillatorplus the
mechanical support and tolerances (one can assume22 to 25 mmtransverse dimensions depending on
themechanicalsupport). Furthermore thelongitudinalend-manifoldlengthisreducedto theminimum
bending radius of the bers (10 cm). The accessibility to the photo-detectorsis better because there
are nomechanicalsupport (likesprings...) in this cornerpartofthe detector. Furthermore in thecase
we introduce a x-yplane coincidence the triggeringsignalsare only traveling between four clusters of
photodetectors/plane.
Thetransportation can bemade separately for the passiveelements (scintillatorsub-modules) and
the photo-detectors boxes. The scintillator and the photo-detector can be tested and calibrated rst
separately on the respective production sites and then connected together with the clear bers for a
globalcalibrationtest. ThistestcouldbeperformedeitherintheGranSassoorinoneoftheproduction
sites.
Thisarchitectureiscompatible withbothplastic andliquid scintillatoroptions. Furthermore itcan
bedesignedforaMaPMToraHPDreadout. Wewillpresentinthisnoteapossibleschemewithanew
ThemainfeaturesoftheHPDpresentedinthisnote(resolution,uniformity,linearity,lowcross-talkand
good quantum eÆciency)havebeenevaluatedin Lyonwith the61-pixelsmodelavailableatthat time.
Since then, DEP has developed for the BTeV collaboration an HPD with 163 pixels. The major
changehasbeenthere-designofthesiliconchip. TheR&Dforthisisnowcompletedandthersttubes
havebeen delivered to BTeV and anotherone couldbe delivered forOPERA in 3month's time. The
design of such a tube is displayed on Fig. 32. It will be a proximity focusing tube with a "hot" S20
photo-cathode.
Figure32: Preliminarydesignofthenew163pixels,proximityfocusingtube.
Thetube is thesame asthe61 pixels (same dimensions, see Fig.1). ThePGA socket is larger (42
mm) and has 192 pins (17 of which are not connected). The pixels have still hexagonal shape with
shorterdimensions(1.345mm atto atinsteadof2mm),1.57mm
2
activeareaand40mgapbetween
adjacentpixels. Fig.33givesthedetailsofthePINdiode. Theoutputcapacitanceisslightlyreduced(3
pFinsteadof4pF).
The rst estimations show that this imposes a maximal diameter of 1.1 mm for the bers. The
cross-talkofsuchacongurationneedsalsoto beinvestigated.
6.2 Proposals for a tracker architecture
Inthebasicdesign 160pixelsofthe HPDareconnected toascintillatorsub-module. These160pixels
correspondto5chipswith32VATA(ortheLALchip)channelseach. Asmentionedbefore,thesolution
iscompatiblewitheitherplasticorliquidscintillator. Inthecaseofplasticscintillator,wecouldsuppose
that westay closeto the baselinedesign with only dierencethe location of thephotodetectors and a
slightreduction of transverse length. We could connect upto 320 strips on four HPD's for one plane
numberof tubes forthe OPERA detector is therefore 576(that is half of thebaseline proposal). The
segmentationofthedetectorcouldbedecreasedbytaking2.1cm widescintillatorstrips.
Inthe case of liquid scintillator the segmentation could be easily varied. We will assume that the
segmentation is11cm
2
(standard polycarbonatecellsalreadytested bythe collaboration). Tokeep
thesamenumberofreadoutchannels,weproposetoreadoutthebersonlyatoneendwithamirror(or
somere ectivematerial)depositedattheotherend. Thenwecouldgather640bersonfourHPD'sfor
oneplane. This requires8HPD'sperx y planejust asin thepreviouscase. Therefore576tubesare
neededinthisarchitecturealso.
Theincreaseinberlengthisbalancedbythelowerpriceoftheliquidscintillator. Inorderto limit
lowsensitivityareas(farawayfromthephoto-detectorand thereforewiththelowest numberof
photo-electronsdueto theabsorptionalongthe ber), thereadout sidemay bealternatedoneachplane. We
show2possiblesketchesforthemodulearchitectureonFig.34. Weseefromtheseschemesthat2HPD's
arehousedinonephoto-detectorboxthatisxedinthecorner. AlightinjectionsystemusingLED'scan
behousedin suchboxes(theconstraintsin termsofweightanddimensionsarereleasedinthecorners).
6.3 Optical connections
AcriticalpointoftheproposedarchitectureistheopticalconnectionfromtheWLSberstothe
photo-detectors. The mosteÆcient wayto performthis connection isto use clearbers which havealonger
attenuationlengththan theWLS bers. Forinstance theMINOS collaborationquotes anattenuation
lengthof14mforthe1.2mmKuraraybers. Thisconnectiongivesthepossibilitytodecouplethe
scin-tillatorsub-modules(passiveelements)fromthereadout(photo-detectorandelectronics). Thedrawback
isthelightlossintheconnector. Neverthelessthe MINOScollaborationhavedesignedconnectorswith
above90% absolutetransmission.
For 90%absolute transmission,thenthemaximal lightlossalongaberstartingfrom thecenter
ofthedetectorandgoingtoaphoto-detectorcookielocatedinacorner(weassume4m length)is30%.
Fortheshorterclearbers(withamaximallengthassumedto be1m)theattenuationis15%.
Twotypesof opticalconnectionscould be foreseen(Fig.35). Therst one(type(a))(MINOS like)
HPD (x2)
HPD (x2)
HPD (x2)
HPD (x2)
mirror
160 clear
fibers
160
160
cells
cells
HPD (x2)
HPD (x2)
HPD (x2)
HPD (x2)
160 clear
fibers
160
cells
cells
80
Figure34: Possiblesketchesforthetrackerarchitecturewithoneortwoendsreadout(resp. leftandrightplot).
Thegureshowsthexandycells. Theclearbersarealsodrawn(greenlines). TheyareconnectedontheHPD
boxeswhichhouse2HPD's160pixelstoxthesegmentation. Intheleftplot,Thehatchedareasrepresentthe
mirrorsattheendofthebers.
with clear bers. This oers theadvantageto have short connectors (with agood relativealignment)
thatshouldbeeasiertoproduce. Inthesecondprocedure(type(b))weuselongbarsinwhichthebers
areglued. TheWLSbersarecutat theexactdimensionofthesub-modules, withoutextralength. As
theabsorptionin thesebersvariesrapidlyatshort distances,weavoidlargespreadsin thelightyield.
Thetotalextralengthofthebersoutofthesub-modulesisalsoreducedw.r.t. therstprocedure. The
weakpointofthisprocedureis theabsolutetransmissionthat couldbeachievedwith such aconnector.
StudiesareunderwayinLyononaliquidscintillatorprototyperealizedincollaborationwithCERN(see
Sec.6.4).
detector
Photo-detector
Photo-WLS fibers
type (a)
type (b)
clear fibers
clear fibers
WLS
fibers
Thepurposeofthisprototypeisthestudyofthelightyieldwiththestandardliquidscintillatorusedby
thecollaboration,thestandard11cm
2
polycarbonatecellsloadedwithTiO
2
andclearberconnection
ofthesecondtype(so-calledtype(b)inthepreviousSection). Ageneralviewoftheprototypeisshown
inFig.36. Itconsistsof60cellsclosedateachendsbya3cmsiliconglueend-capandthersthalfof
anopticalconnector. Eachcellis cutatitsendover1cmin ordertoallowtheliquidtogofrom cell
tocellduringthellingprocedure(seedetailonFig.36).
Figure36: Designoftheliquidscintillator prototype.
Thersthalfoftheopticalconnectorismadeoftwo33cmplastic bars(PVCwill beusedforthe
prototype)thataredetailedinFig.37. Eachbarhas30holesforthebers(wewilluse1.2mmWLSand
clearbers), 30holesonthetopforthebergluing,6alignmentpinsand6screws. One5mmdiameter
holeispiercedintothebarforlling. A5cmtubegoesfromthisholeinsidethedetector(outsideofthe
siliconglueend-cap). Anexternaltubeisusedtoconnecttothellingpump. Oncethellinghasbeen
completed this externaltube isremovedand replaced withadedicated screw forliquid tightness. The
Theupcomingmilestonedoesnotconcernthe triggerandthedata acquisition. Neverthelessthechoice
of a tracking device and detector imposes constraintsdue to singles rates, data loads, dead-timesand
spuriousextractionofbricks.
Furthermore,alargepartoftheLyongrouptestbeamactivityconcernedthevalidationofafulldata
acquisition and readout chain from the photodetector to theworkstation. We analyzethe possibilities
oeredbysuchaschemefortriggeringand readout. Theschemedoesnotdepend onthephotodetector
(MaPMTorHPD)optionanditcanimplementeitherfullautotriggerabilityorax-ycoincidencetrigger.
Ifoneappliessomemodicationstothefrontendthis schemeisalsoapplicableto RPC's.
7.1 Dead time
Intheprevioussections,wepresentedtheVATAcg front-endscheme.
Webrie yrecallitsprinciple: itconsistsof64-channelchipswherethechargeispreampliedandthen
shapedthroughafastandaslowshaper. Theslowshaper(VA)hasarise-timeof1-2microsecondsand
thefastshaper(TAcg)hasarise-timeof70nanoseconds. Thesignalsofthefastshaperpassthrough
adiscriminator,andtheirORservesasaholdsignaltosampleonthepeakthechargeaccumulatedover
the64channelsoftheVApart. Thesamesignalinitiates thereadout sequencewhich clocksthe64fold
multiplexed dataintoanADC. Onethushasafullytrigerlessscheme(schemeA).Thecurrentreadout
speeddeterminedbythemultiplexer(5 MHz) doesnotpermitthe dead-timelessreadoutof a160pixel
HPDbyonlyone ADC
3 .
Thedead-timesofschemeAisgivenbytheformula:
Deadtime=(1 e D ) D=Nr(t p +Nt r )
whereNisthenumberofchannels(e.g64)andristhesinglesrateperchannel.Weassumeashaping
time oft
p
=2microsecondsand a5MHzreadout clock(t
r
=200ns/channel). ForD 1 wehaveof
course (1 e
D
) !D. Figure 39 showsthe percentof deadtime asa function of thesingles rateof
onechannelfor a64foldreadout and 3clock speeds. A nominal 5% deadtime is reached for asingles
rateof 50Hz and5MHzclockreadout. Bydoublingthereadout speedone obtainsthesamelimitfor
100 Hz, and at 20 MHz oneobtains 150 Hz. For comparison the same limit for an optionwhere 160
pixels aremultiplexed in oneADC, and20 MHzreadout is30 Hz. Inconclusion,with thepresentand
forseable readout speeds the full autotriggerability withoutdead time canbe achieved only ifwekeep
the64granularityforthereadoutchain.
Alternativelyone canuse ax-ycoincidence. Wecallthis schemeB. Thecurrentprototypereadout
cardORCAacceptsexternaltriggersandthuspermitstheformationofaglobalbi-plane(XY)coincidence
usingthefast TAORlogicalsignalsofthe16to 20front-ends. The longestlengthalong theperimeter
of one plane is 14m adding a maximum round-trip delay of 100 ns, well below the 2 microseconds
available. In this scheme the TA signal of an x front-end issues a hold, only after validation from a
coincidenceintimewiththecorrespondingy-plane. Thesamehappensforthey-plane. Thisschemehas
theadvantagethat it keepsthepropertimingbetweentheTA and thepeaktime oftheVA, that only
thefront-endswithasignalarereadout,andthatitexhibitsasisshownbelowrobustnesstohighrates.
Thedisadvantageof thisschemeistheneedfor twoextracables comingoutand in fromthe card,and
theneedforanextracoincidenceformingmodule.
Wecalculatetherateofcoincidencesforthisschemebyusingtheformula:
r c =2N 2 r 2 G
where N is the number of strips in a plane and Gis the coincidencegate. For the 2.6cm strip width
optionwehave512channels/planewhileforthe1cmoptionwehave640channels/plane. A100nsgate
Gcanlargelyaccommodatedelaysanddierentdecaytimes.
Figure 39showsthepercent ofdead-timefor a160foldreadout, 1cmwidth stripsread onone side
(640channels/plane)asafunctionofsinglesrate. Wehavethenominal5%deadtimeforsinglesratesof
150,200and250Hzwith 5,10and20 MHzreadoutclockfrequencyrespectively. The160fold readout
(1ADC for160channels)becomesrobustto uctuations insinglesrates.
3
Note that inthe caseof the 160 channel HPD, we would need ve 32channel VATACGchipsthat couldeitherbe
readoutbyasingleADC,orhave2HPD'sreadby3separateautotriggerable64ADCchains;keepingthe64foldelectronics
11
-1
2
3
4
5
6
7
8
9
10
20
30
40
50
60
70
0
50
100
150
200
250
300
350
400
450
500
01/07/15 15.28
Singles rate r (Hz)
DeadTime
%
12
-1
2
3
4
5
6
7
8
9
10
20
30
40
50
60
70
80
90
0
50
100
150
200
250
300
350
400
450
500
01/07/15 15.28
Singles rate r (Hz)
DeadTime
%
Figure 39: Left: %ofDeadtimeas afunction ofasinglechannelrate,inthecase ofa64foldmultiplexingA)
Solidline: 5MHzclock B)Dashedline: 10MHzclockC)Dottedline: 20MHzclock Right: %ofDeadtimeas
afunctionofasinglechannelrate,inthecase ofax-y coincidenceina640channelplaneA)Solidline: 5MHz
clockB)Dashedline: 10MHzclock C)Dottedline: 20MHzclock
7.2 False extractions
Anotherquestion by thereferees is thepossibility of arandom coincidencesimulating aneventtrigger
within thespilltime. Werstapplied2kindsoftriggerforthemostsensitiveevents: !eand !
quasielastic(QE) events. The rsttriggerdemands thecoincidence of at least2 consecutive planes(4
planesin allifonecountsXandY planes),andthesecond demandsasingleplanex-ycoincidencebut
asks for alarge number of pe's (10)and at least 3 stripshit. The eect of theOR of these 2triggers
onthe QEeventsmentionedabovecanbeseenat table 4. ThePlastic and Liquidscintillatoroptions
aretakenwith6and9p.e'sat thecenterrespectively. Thetable intendstoshowthattheabovetrigger
criteriahaveagoodtriggereÆciency.
Condition Liq !e Liq ! Plast !e Plast !
1planecoincidence 0.99 0.995 0.99 0.995
2planescoincidence 0.75 0.93 0.73 0.93
2planesORelectron 0.92 0.975 0.905 0.97
Table4: Trigger eÆciency for dierent scintillatoroptions and the most sensitive QEevents. For the
detailsofthetriggerssee thetext.
Nowthenextquestioniswhether these veryloosecriteria,will induce anunacceptablerateof false
extractions. Thenominaltolerablelevelhasbeensetto0.3bricks/day. Wedidnotevaluatetheelectron
triggerrandomratesinceitdemandsamoredetailedknowledgeofthep.e. spectrum(including
radioac-tivity) that wedo notcurrentlyhave. Thecut of 10p.e is suÆciently high to ensure that we will not
haveanysurprises.
Comingtothe2XYplanecoincidencetrigger,oncanusethefollowingformulatoestimatetherates:
N extr=day =72N 2 [2(23r 2 G) 2 G]t livetime=day
wheretheterm(23r
2
G) calculatestherandomdoublet coincidenceofaxstrip with3adjacent
has to be multiplied with the numberof strip combinations/plane N (N=256 or640),the numberof
consecutiveplanes(72)thetotallive-timewithbeamonperday(220milliseconds).
Onecanseein gure 40thatuptoextremevaluesofsinglesratetheacceptablelimitforfalsebrick
extractionsperdayissatised.
16
-0
0.005
0.01
0.015
0.02
0.025
0.03
0
50
100
150
200
250
300
350
400
450
500
01/07/15 15.28
640/256 channels r (Hz)
Nextr/day
Figure 40: Number of eventsperday that satisfy thecriterium of adouble x-yplane coincidence asa
function ofsinglesrateperchannel. Solid linea640stripplane,Dasheda256stripone.
Inconclusiona100Hzrateisacceptableforfullyautotriggerablerates(schemeA)and64foldreadout,
while plane coincidences (scheme B) can raise the tolerance to 150-200 Hz. These rates are largely