Characterization studies of Silicon Photomultipliers and crystals matrices for a novel time of flight PET detector EtiennetteAuffray3,FaraahBenMimounBelHadj3,DanieleCortinovis∗1,2,KatayounDoroud3,ErikaGarutti2,PaulLecoq3, ZhengLiu4,RosanaMartinez4,MarcoPaganoni4,MarcoPizzichemi4,AlessandroSilenzi1,ChenXu1,2 andMilanZvolský1,2 1DeutchesElektronen-Synchrotron(DESY),Notkestrasse85,22607Hamburg,Germany 2UniversityofHamburg,LuruperChaussee149,22671Hamburg,Germany 3EuropeanOrganizationforNuclearResearch(CERN),1211,Geneva23,Switzerland 5 4UniversityofMilano-Bicocca,PiazzadellaScienza3,20126Milano,Italy 1 0 2 n a Abstract J 7 1 This paper describes the characterization of crystal matrices and silicon photomultiplier arrays for a novel Positron EmissionTomography(PET)detector,namelytheexternalplateoftheEndoTOFPET-USsystem. TheEndoTOFPET-US ] collaborationaimstointegrateTime-Of-FlightPETwithultrasoundendoscopyinanovelmultimodaldevice,capableto t e supportthedevelopmentofnewbiomarkersforprostateandpancreatictumors. Thedetectorconsistsintwoparts: aPET d headmountedonanultrasoundprobeandanexternalPETplate. Thechallenginggoalof1mmspatialresolutionforthe - PETimagerequiresadetectorwithsmallcrystalsize,andthereforehighchanneldensity: 4096LYSOcrystalsindividually s n readoutbySiliconPhotomultipliers(SiPM)makeuptheexternalplate. Thequalityandpropertiesofthesecomponentsmust i beassessedbeforetheassembly. Thedarkcountrate,gain,breakdownvoltageandcorrelatednoiseoftheSiPMsaremeasured, . s whiletheLYSOcrystalsareevaluatedintermsoflightyieldandenergyresolution. Inordertoeffectivelyreducethenoisein c thePETimage,hightimeresolutionforthegammadetectionismandatory. TheCoincidenceTimeResolution(CTR)ofallthe i s SiPMsassembledwithcrystalsismeasured,andresultsshowavalueclosetothedemandinggoalof200psFWHM.The y lightoutputisevaluatedforeverychannelforapreliminarydetectorcalibration,showinganaverageofabout1800pixels h firedontheSiPMfora511keVinteraction. Finally,theaverageenergyresolutionat511keVisabout13%,enoughfor p effectiveComptonrejection. [ 1 v 3 1 Introduction inedwithultrasoundprobesthroughnaturalcavities, 3 thereforethediagnosticcapabilitycanbeenhancedby 2 fusingthemorphologicalimagefromultrasoundwith 4 Positron Emission Tomography (PET) is a non inva- aPETmetabolicimage. Theendoscopicapproachand 0 . sive, diagnostic imaging technique for measuring the theuseoftheTOFinformationwithanunprecedented 1 metabolic activity of cells in the human body. Nowa- CTRof200ps(3cmalongthelineofresponse)allow 0 daysitiswidelyusedforcancerdiagnosis[1]. todefineaspecificRegionOfInterest(ROI)andsignifi- 5 1 Amongthetumors,pancreaticcarcinomaisoneofthe cantlysuppressthebackgroundfromtheneighboring : most aggressive and resistant to current therapies [2], organs. v i and most often is detected only on an advanced state TheEndoTOFPET-USdetectorconsistsofaPETheadex- X ofdevelopment. Ontheotherhand,prostatecanceris tensionmountedonacommercialUSendoscopeplaced r themostcommoncanceramongmales[3]. close to the ROI and an external PET plate facing the a Bothpancreasandprostatearesurroundedbyorgans patient’sabdomen,incoincidewiththePEThead(Fig- with high uptake: the liver and the heart are close to ure 1). The external plate is a square of 23 x 23 cm2, thepancreaswhilethebladderisneartheprostate. In anditismadeof256detectorunitmodules. Eachmod- thesecasesthecurrentfull-bodymultimodalPETscan- uleconsistsina4x4LYSO:Cecrystalmatrixgluedtoa nershavestronglimitations,duetothenoisefromthe discretearrayof4x4analogSiPMsfromHamamatsu. neighboringorgansandthelackofspecificbiomarkers LYSO:Cecrystalshavehighlight yieldandfast decay [4][5]. EndoTOFPET-US[6]aimstoovercometheselim- time, therefore appropriate for TOF PET. SiPMs are its,allowingtoinvestigatetheperformancesofnewly novel photon detectors also suitable for TOF PET ap- developedspecificbiomarkersoftumoralprocessesfor plications [7], thanks to their compactness, high gain, pancreasandprostate. Bothorgansareusuallyexam- ∗Correspondingauthor,[email protected] 1 insensitivitytomagneticfields,lowoperatingvoltage heatgeneratedbytheactivecomponentsontheSiPM andexcellenttimingproperties. motherboardisdissipatedbyafan. Thetemperatureis Thefollowingsectionsdescribetheexperimentalsetups monitoredbyasensor(DallasDS18B20)butnotactively and the results of the characterization of the SiPMs, controlled. crystalsandthecombinedsystem(SiPM+crystal). All the setups have been designed for a fast and reliable measurementofalargequantityofcomponents,rather thanadetailedstudyofsingledevices. Figure2: SketchoftheexperimentalsetupfortheSiPMcharacteri- zation. Figure1: TheEndoTOFPET-USdetectorwithamagnificationof the pancreatic endoscope (top circle) and the external plate(bottomcircle). 2.2 Gain and Breakdown voltage 2 Photodetector Characterization The integrated charge spectrum given by the light pulsesinshowninFigure3. Thisisrepeatedfor30volt- TheSiPMschosenfortheexternalplatearematricesof agesteps,acquiring200keventssampleperstepwith 4x4discreteMPPCs(MultiPixelPhotonCounter)from anintegrationgateof100nssynchronizedwiththelaser Hamamatsuphotonics(S12643-050CN).EachSiPMin pulse. Thefirstpeakcorrespondstothepedestal(zero the matrix is a square of 3.0 x 3.0 mm2 with 3464 ac- pixelfired)whilethefollowingpeakscorrespondtoan tive pixels (50 x 50 µm2 each). The distance between increasing number of pixels fired. After the pedestal two adjacent SIPMs in the matrix (center to center) is subtraction,thespectrumisprojectedintothefrequency 3.6mm. TheseSiPMsexploittheThroughSiliconVia spaceusingaFastFourierTransform. Theperiodofthe (TSV)technology,leadingtolessdeadspaceandare- first harmonic is taken as the value of the SiPM gain ducedconnectionlengthascomparedtoconventional (G)inQDCunits,whichisthenconvertedintonumber wire-bondedSiPM.Thefollowingsectionsdescribethe ofelectronsbythefollowingformula: experimentalsetup,theanalysismethodandtheresults ofthecharacterizationof256SiPMmatrices. G·R qdc G = , (1) 2.1 Experimental setup A The layout of the experimental laboratory setup is showninFigure2. Inalighttightbox,theSiPMmatrix where R is the QDC resolution of 25 fC per QDC qdc ismountedonamotherboardincorporatinglinearam- bin and A is the amplification factor of the readout plifiers(InfineonBGA614)andhighvoltagefilters. The board (19 dB, ∼ 8.9). The error on the gain is about signalforeachofthe16SiPMsofthematrixisobtained 1%, obtained as one bin in the Fourier space plus an using low intensity blue light (Advanced laser diode additionaluncertaintyof5 ,evaluatedbyrepeating system,451nm). TheamplifiedSiPMoutputisreadout themeasurementmultipleti(cid:104)mesunderthesamecondi- by a VME-based charge-to-digital converter (CAEN tions. QDC965A)witharesolutionof25fCperQDCbin. The 2 ×103 Counts 890000 pede1 sptiaxl Gain 11680000 G U=b d(0 =.5 (2634 .3±7 0 .±0 003.0) 2×) 1V06 700 600 2 pix 1400 500 1200 3 pix 400 1000 300 800 200 600 100 0 400 600 800 1000 1200 1400 1600 1800 65.5 66 66.5 67 67.5 68 QDC channel Bias voltage [V] Figure3: Singlephoto-electronspectrumforasingleSiPM,each Figure4: Gainasafunctionoftheappliedbiasvoltageforasingle peakcorrespondstoacertainnumberofpixelfired. SiPM.Theredlineisalinearfitonthedatapointswhose errorsarewithinthedots. Thebreakdownvoltage(U ) bd andgain(G)at1Vexcessbiasobtainedfromthefitare givenintheinlet. Figure4showsthegaindependenceonthevoltage biasapplied. Asexpected,thegainislinearlypropor- tionaltothebiasvoltageapplied,accordingto G =C ·∆U =C (U −U ), (2) pixel pixel bias bd M 400 EEnnttrriieess 44009966 P where Cpixel is the pixel capacitance, ∆U is the excess Si 350 MMeeaann 00..44882266 bisiatshevoblrtaeagke,dUowbians ivsotlhtaegaep. pHlieendcbeiaUs voisltaogbetaainneddUbbyd # of 300 RRMMSS 00..0011449922 bd 250 applyingalinearfitandextrapolatingtozerogain,with 200 anuncertaintyoftheorderofafewtensonmV.Figure 5showsthedistributionofthegainslope(G at1Vex- 150 cessbias)foralltheSiPMs. Themeanvalueis0.48·106 100 V−1 withaspreadof3%amongtheSiPMstested. 50 It is known that the Ubd depends on the temperature 00.35 0.4 0.45 0.5 0.55 0.6 [8], and some fluctuations have occurred during the Gain slope [106 V-1] wholesetofmeasurements. Moreover,localdifferences in temperature have been observed within the SiPM Figure5: Distributionofthegainslope(gainpervolt)forallthe matrix, due to the inhomogeneous dissipation of the SiPMs. heatgeneratedbytheamplifiers. Thereforeanoffline temperature correction (see section 2.5) has been ap- plied in order to compare the results. Figure 6 shows thedistributionofthebreakdownvoltagesat25 ◦Cfor alltheSiPMs. AlthoughtheU spreadforalltheSiPMisabout2V, M 140 EEnnttrriieess 44009966 bd P thismustnotexceed0.5VineachsingleSiPMmatrix, Si 120 MMeeaann 6644..2299 becausethisisthemaximumvoltagebiastuningrange of RRMMSS 00..44333344 # 100 of the ASIC developed for the SiPM readout [9] [10]. AsshowninFigure7,alltheSiPMmatricesrespectthis 80 requirement. 60 Finally, the results obtained have been compared to 40 theoperatingvoltages(U )providedbytheSiPMpro- op ducer. TheU isdefinedas: G(U ) =1.25·106. The 20 op OP differencesbetweentheU calculatedusingequation 0 op 63 63.5 64 64.5 65 65.5 66 2withdatafromthemeasurementsandUop provided Breakdown voltage [V] byHamamatsuareshowninFigure8. Theagreement isgood,despiteashiftof6%andaspreadof10%. This Figure6: Distributionoftheextractedbreakdownvoltageforall isduetothedifferentmeasurementmethodimpliedby theSiPMs. theSiPMproducerandtheuncertaintyofU and G. bd 3 events n is given by n = DCR ·∆t. The number 0.5pix ay EEnnttrriieess 225566 of events in the pedestal distribution is given by the arr 25 MMeeaann 00..11440011 Poissonprobabilitydensityfunctionforzero DCR0.5pix M RRMMSS 00..0055776655 eventsinthetime ∆t: P 20 Si # of 15 P (∆t) = e−DCR0.5pix·∆t. (3) 0 10 5 The DCR isthereforecalculatedusing: 0.5pix 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Ubdmax - Ubdmin [V] DCR = ln(N0/Ntotal), (4) 0.5pix ∆t Figure7: Distributionofthemaximum∆U ,i.e. thedifference bd between maximum and minimum U in one matrix. bd Thedeviationisbelowthe0.5Vrequirementspecifiedto where N isthenumberofeventsinthepedestalpeak 0 theSiPMproducer. and N isthetotalnumberofeventsrecorded. The total erroronthecalculatedvaluedependsonthestatisticof thesampleandonthequalityofthespectrum. ForeachSiPMthe DCR isevaluatedfor30voltage 0.5pix levels, in a range of 3 V starting about 1 V above the M EEnnttrriieess 44009966 breakdown voltage. 500k event samples are collected P Si 300 MMeeaann 00..0066445566 foreachvoltagepoint. of RRMMSS 00..11009944 Also the DCR has a dependence on the temperature, # 250 thereforeacorrection(seesection2.5)hasbeenapplied 200 inordertocompensatefortemperaturevariations. 150 The distribution of DCR0.5pix, corrected to 25 ◦C and at 2.5 V excess bias, is shown in Figure 10 for all the 100 SiPMs. 50 TheDCRhasanegativeimpactonthetimeresolution. 0 The best time resolution is achieved when the timing -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 (Uproducer - Umeasurement) [V] thresholdonthereadoutASICissetaslowaspossible, op op uptothenoiselevel. Thereforea DCR acceptance 0.5pix Figure8: DifferencebetweenUop givenbytheSiPMproducerand limitof3MHzforeachSiPMhasbeenagreedwiththe Uop calculatedfromthemeasurement. SiPM producer. Almost all the SiPMs are within the limit, however 5 matrices have been sent back to the producerbecausetheyincludedatleastoneSiPMwith excessive DCR . 0.5pix 2.3 Dark Count Rate Eveninabsenceoflight,anavalanchebreakdowncan betriggeredbyaphotoelectrongeneratedfromthermal s excitation or by tunneling effect. This phenomenon, unt 16000 pedestal usuallycalledDarkCountRate(DCR),isPoissondis- o C 14000 tributed and it is measured by randomly integrating thechargewithatimewindow∆tof100ns(sametime 12000 10000 windowasforthegainmeasurement). Atypicalnoise 8000 spectrumisvisibleinFigure9. 0.5 pix threshold 6000 Theprobabilityofhavingtwodarkeventsinthetime ∆t is very low, therefore the events corresponding to 4000 1 pix more than 1 pixel fired are mainly due to correlated 2000 noise,whichwillbedescribedinthenextsection. 0450 500 550 600 650 700 750 800 The events in the pedestal peak, i.e. the events with QDC channel chargelessthan0.5pixelthreshold,arenotaffectedby correlated noise, hence they are used to calculate the Figure9: DCR charge spectrum. The events below the 0.5 pix uncorrelatedDCR,definedas DCR0.5pix. threshold(pedestalpeak)areusedtocalculateDCR0.5pix. Inatimerange ∆t, theexpectednumberof DCR 0.5pix 4 M EEnnttrriieess 44009966 M EEnnttrriieess 44009966 SiP MMeeaann 11..444411 SiP 400 MMeeaann 2299..1122 of 102 RRMMSS 00..33667788 of 350 RRMMSS 11..882299 # # 300 250 200 10 150 100 1 50 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 10 15 20 25 30 35 40 45 50 DCR [MHz] Correlated noise probability [%] Figure10: DistributionoftheDCR0.5pix at2.5Vexcessbiasand Figure11: CorrelatednoiseprobabilityPcn at2.5Vexcessbias. at25oCforalltheSiPMs. Theredlineindicatesthe threshold of acceptable DCR. Entries marked in red correspondtotherejectedSiPMs. 2.5 Temperature dependence As mentioned in previous sections, SiPM properties dependontemperature. Therefore,DCR,gain,break- 2.4 Correlated noise downvoltageandcorrelatednoisehavebeenmeasured atdifferenttemperaturesforasingleSiPMinanarray. Usingaclimatechamber(EspecLU-123),severalvolt- Correlated noise includes after-pulse and inter-pixel agescanshavebeenperformedinatemperaturerange opticalcrosstalk[11]. between 6 ◦C and 30 ◦C. Figure 12 shows the linear After-pulsesaregeneratedifanelectronproduceddur- temperaturedependenceofthebreakdownvoltage;the inganavalancheistrappedandreleasedatalatertime, uncertainty on the temperature is included but it is with a delay ranging from nanoseconds up to several smallerthanthemarkersize. Thecoefficientobtained microseconds. Depending on the trapping time con- from the linear fit is 70.1 mV/◦C, and is used for the stantandthepixelrecoverytime,trappedelectronscan temperaturecorrectionsappliedinsection2.2and4.1. generateanavalancheofequivalentorsmallercharge The dependence of the DCR as a function of the tem- withrespecttotheDCRpulses. perature should scale with the density of the thermal Inter-pixelcrosstalkisduetoopticalphotonsgenerated carriers[12]accordingtoequation: duringapixelbreakdown. Thesephotonshaveacertain probabilitytoreachtheneighboringpixels,triggering newavalanches. n(T) =CpixelT3/2e−kEBaT, (6) The precise measurement of each of the two effects is beyondthescopeofthisstudy. However,it’spossible toestimatethecorrelatednoisetriggeringprobability whereC isthepixelcapacitance, T isthetempera- pixel Pcn as: tureoftheSiPMinKelvin, Ea istheactivationenergy and k istheBoltzmannconstant. B Figure13showstheDCRasafunctionoftemperature DCR at2.5Vexcessbias. Anactivationenergy E =0.6eVis 1.5pix a P = , (5) cn DCR obtainedbyfittingthedatawithequation6. Thisvalue 0.5pix iscompatibletopreviousstudies[13],andithasbeen usedforthetemperaturecorrectionsappliedinsection 2.3. where DCR iscalculatedbysettingthethreshold Thechangeinthegainisfoundtobelessthan1%/10◦C, 1.5pix to1.5pixelfired. Figure11showsthecorrelatednoise which is negligible for the temperature variations oc- probability at 2.5 V excess bias for all the SiPMs. The curredduringthegaincharacterizationofalltheSiPMs. averageresultsobtained,about30%,iscompatiblewith Alsothecorrelatednoiseshowsamilddependenceon measurements on similar devices and does not repre- thetemperature,butitisnotrelevantforthedetector sentanissueforthedetectoroperation. performances. 5 theenergyresolution. Figure14showsthedistribution V] oflightyieldobtainedwithMiniACCOSsetupforall [Ubd 6644..68 ∆U/∆T = 70.1 mV/ oC the256matrices. Thedistributionspreadis6.4%,which subtractingtheuncertaintyofthesetupgivesa4.3%as 64.4 lightoutputdispersion. Figure15showsthedistribu- 64.2 tionoftheenergyresolutionforallthemodules,with 64 a mean value of 13.4%. The spread is comparable to 63.8 thebenchuncertainty,thereforethevariationsinenergy 63.6 resolutonarejustduetothesetup. 63.4 In the final configuration, each matrix will be glued 63.2 to a SiPM array. The absolute values of light output 63 5 10 15 20 25 30 obtainedwiththeMiniACCOSsetuphavethereforeto Temperature [ oC] be rescaled to this configuration. In order to obtain this,lightoutputforasubsetof13matriceswasmea- Figure12: Breakdownvoltagemeasuredatdifferenttemperatures. suredonastandardXP2020QPMT,usingRhodosil47V Theredcurveisthelinearfittothedatapoints. The greaseasopticalcouplingandteflonasbackwrapping. coefficientobtainedformthefitisdisplayedintheinlet. Thescalingfactorbetweenthismeasurementcondition and MiniACCOS was found to be 13.15, with a corre- lationcoefficientof0.89betweenthetwodatasets,as showninFigure16. Theaveragelightyieldforallthecrystalmatricescan Hz] 1.4 thereforebederivedas32500 ± 2700Ph/MeV,includ- M [pix 1.2 Ea = 0.6 eV itnemgaintictshaeriesvinaglufartoiomnthofetchaeliburnatcieorntapinrotyceaslss.oBathseedsyosn- 5 1 R0. thestudiesreportedin[15],itcanbeestimatedthatat C 0.8 least20000 − 25000Ph/MeVarenecessarytoachievea D 0.6 CTRof200psFWHM,whichmakeoursetofmatrices perfectly suitable for this purpose. The homogeneity 0.4 on the crystals performance is also ensured within a 0.2 4.3% level for all 256 matrices tested with an energy resolutionof13.4 ± 1.3%asmeanvalue. 5 10 15 20 25 30 Temperature [ oC] Figure13: DCR calculatedatdifferenttemperatures. Thered 0.5pix lineisthefittodatapointsusingequation6. Thevalue ofEa obtainedfromthefitisdisplayedintheinlet. 3 Crystals characterization The crystals for the external plate are grouped in 256 matrices of 4x4 LYSO:Ce scintillators, produced by Crystal Photonics Inc. The volume of each crystal is es 18 EEnnttrriieess 225566 3.5x3.5x15 mm3 and within the matrix they are sepa- ric 16 MMeeaann 22550044 at RRMMSS 115599..99 rated by a reflector foil (ESR Vikuiti by 3M). In order m 14 toensurethesystemuniformity,measurementsoflight of 12 # outputandenergyresolutionhavebeenperformedon 10 bothfacesforeachmatrixusingtheMiniACCOSsetup 8 [14], which allows a fast, automatic and highly repro- 6 ducible data acquisition process. In this setup, each 4 matrix is placed on a custom made teflon plate, for 2 a total of 25 matrices per plate. An air gap of 5 mm 10000 1500 2000 2500 3000 3500 4000 is left between the matrix face and the PMT window. LY [Ph/MeV] Systematicmeasurementswereperformedtoevaluate uncertainties arising from the bench, yielding a total Figure14: Lightyielddistributionsforallthe256crystalmatrices. relativeerrorof4.7%forthelightoutputand12.8%for 6 es 18 EEnnttrriieess 225566 ric 16 MMeeaann 1133..4444 at RRMMSS 11..332244 m 14 of 12 # 10 8 6 4 2 0 10 11 12 13 14 15 16 17 18 19 20 Energy Resolution [%] Figure15: Energyresolutiondistributionsforallthe256crystal matrices. V] Figure17: Thebenchforthemoduleassembly. Thebinocularwith h/Me45000 χχχχ2222 //// nnnnddddffff 3333....666699999999 //// 11112222 thecameraisplacedoverthe5supportsforthecrystal T [P pppp0000 11113333....11115555 ±±±± 0000....2222555588884444 matrices, the SiPM matrices (not visible) are located M _P40000 below. Y L 35000 4.1 Detector light output 30000 The detector light output for the 511 keV photon (de- 25000 fined here as LO ) is of primary importance for the 511 detectortimeresolution. Asdemonstratedinotherstud- 20000 ies [15], the number of detected photoelectrons must 2000 2200 2400 2600 2800 3000 bemaximizedinordertoachievethebesttimeresolu- LY_MiniACCOS [Ph/MeV] tion. Moreover,allthechannelsshouldhavethesame response for the 511 keV photon, allowing an easier Figure16: CorrelatedMiniACCOS-PMTdataandthelinealfit- tingfunctionwithits95%confidenceintervalband. detectorcalibration. A22Na sourcewithanactivityofabout1MBqisused asaβ+emitter. Thesetupissimilartotheonedescribed insection2.1buttheSiPMsignalisnotamplifiedand 4 Detector Module Characteriza- theresolutionoftheQDCissetto200fC.Thecharge integrationgateissetto450nsanditistriggeredbythe tion SiPMsignalitself,usingadiscriminator(CAENmodel 96)andagategenerator(LeCroymodel222). Achan- Adedicatedbenchhasbeendevelopedtogluethecrys- nel switcher (Keithley 7002 switch system) provides talmatricesontheSiPMarrays,allowingtoassemble5 automaticscanthroughthe16channelsineachmatrix. modulesperday. Figure17showsthesetup: thecrystal Thetemperatureisrecordedbeforeeachchannelmea- matricesareplacedonamovablesupportandacamera surement. EachSiPMisoperatedatU asdefinedin op allowsthevisualalignmentwiththeSiPMmatriceslo- section 2.2, eventually corrected for temperature vari- catedunderthecrystals. Theamountofglue(RTV3145) ationsaccordingtothecoefficientof70.1mV/◦C(see iscalculatedinordertohaveamaximumthicknessof section2.5). 0.1 mm, and it is spread over the SiPM matrix with a Fig 18 shows a typical 22Na spectrum for a single dispenser. TheSiPMmatrixisthenpushedagaintsthe channel. Thepeakcorrespondingtothe511keVphoton crystalsandavisualinspectionisperformedtocheck inthechargespectrumisfittedbyaGaussianfunction. ifanyairbubbleoccurred. Ifnone,themodulesareleft Hencethelightoutput LO ,expressedinnumberof 511 onthesupportforthecuringtime(24h). pixels fired on the SiPM, is obtained by the following Everymoduleisthenmeasuredregardinglightoutput, relation: energyresolutionandCTR,whicharedescribedinthe (Q−P)·r qdc nextsections. LO511 = , (7) G 7 where Q is the mean value of the Gaussian fit, r is 450nsintheQDC.Thechannelswitcherprovidesthe qdc 200 fC per bin, P is the pedestal position, and G is automaticscantroughthe15channels,eachoperated 1.25x106. The uncertainty on LO , of the order of atitsownU . 511 op about 30 pixels, is given by statistical error and tem- Ineverychannel,onlytheeventscorrespondingtothe peratureuncertaintyof0.5 oC.NoerrorfortheU is photopeakinthetriggerareselected. Therefore,divid- op available. ingthemeanchargeoftheseeventsbythephotopeak The light output distribution for all the channels is position in the trigger, the cross-talk for the 511 keV shown in Figure 19, the average value of about 1800 photon is obtained for all the remaining 15 channels pixelsshouldallowtoreachtheCTRof200psFWHM. inthemodule. Themeasurementisrepeatedbyselect- A preliminary detector calibration will be based on ingdifferenttriggerchannelsinthemodule. Figure20 thesemeasurements. showstheaveragecrosstalkobtainedfrom4different measurements: everytimeoneofthecentralchannels inthemoduleisusedasatrigger. Thecross-talkdoes nts not exceed 20%, and can be suppressed by setting a ou 500 thresholdat100keV. C 400 300 200 100 0 1000 1500 2000 QDC channel Figure18: 22Na spectrumforonechannel. Thehigherpeakcor- respondstothefullabsorptionofthe511keVphotons, while the peak on the right comes from the 1277 keV Figure20: Averagecross-talkforthe511keVphotonobtainedover gammaemittedduringthede-excitationofNeontothe 4measurements,usingeachtimeacentralchannelas groundstate. trigger. 4.3 Energy calibration els 350 EEnnttrriieess 44009966 It is well known that the SiPM is a non-linear device nn 300 MMeeaann 11882233 due to the limited number of pixels, therefore a non- a RRMMSS 111199..11 h linear correction is necessary in order to estimate the c 250 r energyresolutionforthe511keVgamma. o ct 200 Using the setup described in section 4.1, the detector e et 150 lightoutputismeasuredfordifferentgammaenergies: d of 100 356keVfrom133Ba,511and1277keVfrom22Na,and r. 662keVfrom137Cs. N 50 Themeasurementisperformedonasampleof4mod- 0 ules (64 channels): for each channel the light output 1000 1500 2000 2500 3000 LO [pixel fired] data are fitted with the function describing the SiPM 511 saturation: Figure19: Lightoutputdistributionforallthechannels. Npix = Nmax(cid:16)1−e−N(cid:101)m·Eax(cid:17), (8) where N is the average number of pixels fired on 4.2 Inter channel cross-talk pix thephotodetector, (cid:101) isthenumberofpixelfiredonthe The cross-talk between the module channels is mea- SiPMperunitofenergydepositedinthecrystaland E sured on a sample module. Similarly to the setup de- is the energy of the detected gamma photon. N is max scribedinsection4.1,the22Na sourceisplacedinfront relatedtothemaximumnumberofpixelsintheSiPM. ofthemoduleandonlyonechannelischosenastrigger. Thisnumberisactuallyhigherthanthenominalvalue Anyotherchannelinthemoduleisacquiredsimultane- (3464 pixels) because the pixel recovery time is about ouslytothetrigger,usingacommonintegrationgateof 20ns,whichishalfofthedecayconstantofthecrystal 8 (40ns). Figure21showsthenon-linearfunctionforonechannel, x a m howeverthisfunctionisnotuniqueforallthechannels. N 5800 It is expected that the parameter (cid:101) varies among the 5600 channelsaccordingto LO ,whichtakesintoaccount 511 thevariationsintheSiPMproperties(photodetection 5400 efficiency,correlatednoise),crystallightyieldandthe 5200 opticalcoupling. Thedependenceof Nmax and (cid:101) tothe LO511 forthe64 5000 χχ22 // nnddff 2277..5566 // 6622 channelsisshowninFigure22and23respectively. Al- pp00 11..114477 ±± 00..22881199 4800 pp11 33113399 ±± 553388..99 thoughthelineardependenceof (cid:101) wasexpected,also N showsadependenceon LO ,maybeduetovari- 1700 1800 1900 2000 2100 max 511 LO [pixel fired] ationsininthepixelrecoverytimeorcrystaldecaytime. 511 In both cases a linear fit is applied and the relations obtainedareusedtocalculate N and (cid:101) foranyother Figure22: Distribution of the parameter Nmax of equation 8 de- max pendingonLO for64channels. Theredlineisthe channeldependingonits LO . 511 511 linearfit. Figure 24 shows the energy obtained using equation 8 and the LO for every channel. Similarly, the cor- 511 responding energy of the 1277 keV peak of the 22Na spectrumisestimatedforeverychannel(Figure25). In bothcasesthepredictedenergyisingoodagreement withthemeasurements,theprecisionworsensonlyif ∈ 5.4 χχ22 // nnddff 2222..5599 // 6622 the channel LO511 is significantly lower than the aver- 5.2 pp00 00..000022883377 ±± 00..00000011338844 age. pp11 00..88115511 ±± 00..22662288 5 Finally, after the energy calibration has been applied, 4.8 theenergyresolutionforthe511keVgammaisevalu- 4.6 atedforallthechannels(Figure26). Onlyonechannel isoutofthedetectorrequirement: atleast20%isnec- 4.4 essarytoeffectivelydiscriminatethe511keVgammas 4.2 fromtheComptonevents. 4 3.8 1700 1800 1900 2000 2100 LO [pixel fired] 511 Figure23: Distributionoftheparameter(cid:101)ofequation8depending onLO for64channels. Theredlineisthelinearfit. 511 el EEnnttrriieess 44009966 n 200 n MMeeaann 551144..99 x 5000 ha 180 RRMMSS 11..550022 Npi 44050000 χχNN∈∈22 mm //aa xxnn ddff 44.. 55 5555441100 ±±88 88 00±±..55.. 00662288330088 33//00 ..335566 ector c 111246000 3500 et 100 3000 of d 80 2500 # 60 2000 40 1500 20 1000 0 500 505 510 515 520 525 530 500 Energy [keV] 0 0 200 400 600 800 1000 1200 1400 energy [keV] Figure24: EnergyDistributioncalculatedfromthepositionofthe 511keVpeak(from22Na)forallthechannelsafterthe Figure21: Detectorlightoutputfordifferentgammaenergies,the energycorrectionhasbeenapplied. Thetailontheright dataarefittedwithequation8. correspondtothechannelswithverylowLO . 511 9 Figure30showstheCTRasafunctionofbiasvoltage el 200 EEnnttrriieess 44009966 fortwochannelsincoincidencewiththereferencemod- n n 180 MMeeaann 11228822 ule. DuetotheincreaseoftheSiPMPhotonDetection a h 160 RRMMSS 1166..5511 Efficiency (PDE) with the increase of excess bias, it is c or 140 expectedthatalsotheCTRimprovesaccordingly. How- ect 120 ever, also the DCR increase with the excess bias, and det 100 at a given point it becomes predominant. In order to of 80 comparetheCTRofallthe256modules,anexcessbias # 60 of 2.5 V is used for any channel, the temperature is 40 fixed at 19 oC and the NINO ASIC threshold is kept 20 fixedat55mV,whichisequivalenttothethresholdof 0 1200 1220 1240 1260 1280 1300 1320 1340 1360 1380 1400 0.5photoelectron. TheCTRdistributionforallthechan- Energy [keV] nels,measuredwithrespecttothereferencemodule,is presentedinFigure31andshowsanaverageof239 ± Figure25: EnergyDistributioncalculatedfromthepositionofthe 1277keVpeak(from22Na)forallthechannelsafterthe 10psFWHM. energycorrectionhasbeenapplied. nel 103 EEnnttrriieess 44009966 n MMeeaann 1133..2288 a h RRMMSS 00..55338855 c or 102 ct e et d of 10 # 1 Figure27: Thesetupusedforthetimingmeasurements. Thetwo 0 5 10 15 20 25 Energy resolution (@ 511 keV) [%] detector modules are placed facing each other, 10 cm apart.A22Nasourceisplacedbetweenthetwomodules. Figure26: Energyresolutionat511keVforallthechannels. 4.4 Coincidence Time Resolution The coincidence time resolution of all the channels is measuredusingthesetupshowninFigure27. Thetwo w[8] {w[5]>200000} modulesarefacingeachotherat10cmdistance,while Counts330500 CCoonnssttTTaannttoo TT11 229988..33 a22Napointlikesourceisplacedbetweenthemonthe 250 MMeeaann 22..223388ee++0055 SSiiggmmaa 44779944 200 line connecting the module centers. One module has 150 beenselectedasthereferenceforthemeasurementof 100 theother256modules. 50 TheoutputsignaloftheSiPMisamplifiedanddiscrim- 0100 120 140 160 180 200 220 240 260×103 ToT (ps) inatedbyNINOASIC[16]andsenttoahighprecision w[5] {w[8]>200000} TDC(HPTDC-25psLSB)[17]. Thedataanalysisisper- Counts330500 CCoonnssttTTaannttoo TT22 229977..77 formedwithatotalnumberof4x106triggers. Asshown 250 MMeeaann 22..224433ee++0055 SSiiggmmaa 44662277 inFigure28,theTime-over-Threshold(ToT)spectraare 200 obtained for any two channels in coincidence; then 150 100 a gaussian fit is applied to the peaks corresponding 50 to the 511 keV gamma events. Hence the time differ- 0100 120 140 160 180 200 220 240 260×103 ToT (ps) ence for the gamma events of the two channels are histogrammedandshowninFigure29. AGaussianfit Figure28: TypicalTimeoverThresholdspectrafortwochannels is applied and the FWHM is taken as the CTR of the incoincidence. channelmeasured(227.3ps ± 3.2psinthiscase). 10