The GlueX Start Counter Detector
Eric Pooser, Fernando Barbosa, Werner Boeglin, Charles Hutton, Mark Ito, Mahmoud Kamel, Puneet Khetarpal, Anthony Llodra, Joseph Sandoval, Simon Taylor, Timothy Whitlatch, Stephanie Worthington, Carlos Yero, Benedikt Zihlmann
TThe
GlueX
Start Counter Detector
E. Pooser a,b, ∗ , F. Barbosa a , W. Boeglin b , C. Hutton a , M.M. Ito a , M. Kamel b , P. Khetarpal b , A. LLodra b ,N. Sandoval a , S. Taylor a , T. Whitlatch a , S. Worthington a , C. Yero b , B. Zihlmann a a Thomas Jefferson National Accelerator Facility, Newport News, VA 23606, USA b Florida International University, Miami, FL, 33199, USA
Abstract
The design, simulation, fabrication, calibration, and performance of the
GlueX
Start Counter detector isdescribed. The Start Counter was designed to operate at integrated rates of up to 9 MHz with a timingresolution in the range of 500 to 825 ps (FWHM). The Start Counter provides excellent solid angle coverage,a high degree of segmentation for background rejection, and can be utilized in the level 1 trigger for theexperiment. It consists of a cylindrical array of 30 thin scintillators with pointed ends that bend towardsthe beam line at the downstream end. Magnetic field insensitive silicon photomultiplier detectors were usedas the light sensors.
Keywords:
GlueX , Multi-Pixel Photon Counter, Plastic Scintillator, Silicon Photomultiplier
1. Introduction
The
GlueX experiment, staged in Hall D atthe Thomas Jefferson National Accelerator Facility(TJNAF), primarily aims to study the spectrum ofphoto-produced mesons with unprecedented statis-tics in search for gluonic degrees of freedom. Thecoherent bremsstrahlung technique is implementedto produce a linearly polarized photon beam thatimpinges on a liquid H target. A Start Counter de-tector was fabricated to properly identify the pho-ton beam buckets and to provide accurate timinginformation.
2. Design
In this section we discuss the details of the
GlueX
Start Counter design including the scin-tillators, support structure, light sensors and readout electronics.
The Start Counter detector (ST), shown in Fig. 1,surrounds a 30 cm long liquid H target while pro-viding ∼
90% of 4 π solid angle coverage relative tothe target center. The primary purpose of the ST ∗ Corresponding Author
Email address: [email protected] (E. Pooser)
Scintillator PaddlesRohacell Support StructureTarget ChamberLH2 TargetSiPM pre-Amplifier and Bias Voltage Distribution Tedlar Cover
Figure 1: The
GlueX
Start Counter mounted to the liquidH target assembly. The beam direction is oriented from leftto right down the central axis of the ST. is to identify the photon beam bucket associatedwith the event reaction. It is designed to operateat tagged photon beam intensities of up to 10 γ/ sin the coherent peak where the photons range inenergy from 8.4 to 9.0 GeV[1]. The ST has a highdegree of segmentation to limit the per paddle rateswhile also providing background rejection informa-tion. In order to resolve the 4 ns electron beambunch structure delivered by the CEBAF to Hall-Dwith 6 σ accuracy, the GlueX
Start Counter time
Preprint submitted to Nuclear Instruments and Methods in Physics Research A January 10, 2019 a r X i v : . [ phy s i c s . i n s - d e t ] J a n esolution is required to be <
350 ps. It also facil-itates particle identification and can be utilized inthe level 1 trigger of the
GlueX experiment duringhigh luminosity running[2][3].The ST has a cylindrical shape consisting ofan array of 30 scintillators. Their pointed endsbend towards the beam line at the downstream end(Fig. 2). EJ-200 scintillator material from Eljen
Figure 2: 2-D cross section of the Start Counter.
Technology[4] was selected for this application. EJ-200 has a decay time of 2.1 ns and a long attenu-ation length[5]. Silicon photomultiplier (SiPM) de-tectors were selected as the light sensors. Thesesensors are not affected by the 2 T magnetic fieldproduced by the
GlueX superconducting solenoidmagnet. The SiPMs were placed as close as possi-ble to the upstream end of each scintillator elementthereby minimizing the loss of scintillation light[2].
Individual paddles were machined from long,thin, scintillator bars. Each paddle was manufac-tured to be 3 mm thick and diamond milled to be600 mm in length and 20 ± . ◦ arc of radius120 cm following the straight section. The taperednose region is located past the target chamber andbends towards the beam line such that the tip ofthe nose is at a radial distance of 2 cm from thebeam line. . o R . o Figure 3: Start Counter single paddle geometry. Unlabeleddimensions shown are in mm.
After the straight scintillator bar was bent to thedesired geometry the two flat surfaces, oriented or-thogonal to the wide top and bottom surfaces, werecut at a 6 ◦ angle. During this process, the widthof the top and bottom surfaces of the straight sec-tion were machined to be 16.92 mm and 16.29 mmwide respectively. Thus, each of the paddles maybe rotated 12 ◦ with respect to the adjacent paddlesso that they form a cylindrical shape with a conicalend. This geometrical design for the ST increasessolid angle coverage while minimizing multiple scat-tering. The ST scintillator paddles are placed atop a lowdensity Rohacell[7] foam support structure whichenvelopes the target chamber, illustrated in Fig. 1.The Rohacell is 11 mm thick and is rigidly attachedto the support hub at the upstream end and ex-tends along the length of paddles, partly coveringthe conical nose section. The cylindrical part ofthe Rohacell is reinforced with three layers of car-bon fiber, each with a thickness of 650 µ m; this isillustrated in green in Figures 2 and 4.The various layers of material that comprise theST are illustrated in Fig. 4. To ensure that thedetector is light-tight, a plastic collar was placedaround the top of the SiPMs at the upstream end.The collar serves as a lip to which a cylindrical sheetof light insulation film (Tedlar[8]) is attached. Thenose section is covered by a cone of Tedlar whichconnects to the cylindrical section. An additionalcone of Tedlar is taped to the nose of the Rohacelland attached to the top Tedlar cone layer in order toensure light-tightness. A summary of the materialsutilized in the ST are presented in Table 1.2 tem Brand name Material Thickness (mm) Density ( g / cm )Carbon fiber support Carbon fiber Carbon 1.950 1.523
Rohacell support
Rohacell Polymethacrylimide 11.0 0.075
Radial shims
Kapton Polyimide (type HN) 0.762 1.42
Reflective film
Aluminum foil Aluminum 0.016 2.70
Scintillator
EJ-200 Polyvinyltoluene 3.0 1.023
Bundling wrap
Stretch film Polyethelene 0.101 0.917
Light insulation film
Tedlar Polyvinyl Fluoride 0.050 1.50
Table 1: Start Counter materials.
ScintillatorsSiPM Array Rohacell Support TedlarCarbon Fiber Support Light Tightening CollarSupport Hub
Figure 4: Start Counter materials.
Each scintillator bar is read out with an array offour magnetic field insensitive Hamamatsu S10931-050P multi-pixel photon counters (MPPCs)[9].Studies of several photo-detectors were performedin the initial design phase of the ST[10]. Based onthese studies, the S10931-050P model was selected.An individual 3 × MPPC, here referred to asa “SiPM”, consists of 3600 individual 50 × µ m avalanche photo-diode (APD) pixel counters oper-ating in Geiger mode. The signal output from eachSiPM is the sum of the outputs from all 3600 APDpixels[11].The SiPM detectors are housed in a ceramiccase that is surface mounted to a custom-fabricatedprinted circuit board (PCB). The PCB is rigidly at-tached to the lip of the upstream support hub. Theindividual ST scintillators are coupled to the SiPMarrays via an 250 µ m air gap. There are three primary components of the STdetector readout system. The first component “ST1”, shown in Fig. 5, collects scintillation lightfrom three paddles independently and distributesthe bias voltages for the SiPMs. Each array of four
Figure 5: ST1 of Start Counter readout system. Only thecentral array is populated with SiPMs. Approximately 72%of the scintillator light is collected at the upstream end. TheST readout has 10 ST1 units in total. The ruler shown aboveis in inches.
SiPMs has a thermocouple for temperature moni-toring.The second component “ST2”, shown in Fig. 6,has three pre-amplifiers, three buffers, and threefactor-five amplifiers. The output of each preampis split; buffered for the analog-to-digital converter(ADC) output, and amplified for the time to dig-ital converter (TDC) output. The ADC outputsare digitized by Jefferson Lab VXS 250 MHz FlashADC modules[12]. The TDC outputs are inputinto Jefferson Lab leading edge discriminators, fol-lowed by a high resolution 32 channel JeffersonLab VXS F1TDC V2 module[13]. Furthermore,the ST2 has three bias distribution channels withindividual temperature compensation via thermis-tors. The ST2 is attached to the ST1 via a 90 ◦ hermaphroditic connector.The third component of the readout system,“ST3”, provides the interface for the power and3 igure 6: Fully assembled ST readout system. The ST2 unitis connected behind the ST1. The full readout system iscomprised of 10 ST2 units. bias supplies. It also routes the ADC and TDCoutputs as well as the thermocouple output. TheST3 is installed upstream of the Start Counter nextto the beam pipe. A schematic of the ST readoutelectronics is illustrated in Fig. 7. Scintillators (30) ST1 SiPMs (120)ST2
SiPM PCB Readout (10)JLab 250MHz
Flash ADC (2)
RG58
JLab
Leading Edge
Discriminator (2) JLab
F1TDC V2 (1)
RG58 ECL
ISEG
EHS 201P-F-K
SiPM Bias Supply
D-SUB 44
MPOD
MPV 8008
8V / 5A
Low Voltage
Power Supply
D-SUB 37
ST3
Power & Signal Distribution Interface
Figure 7: Start Counter readout electronics diagram. Num-bers in parenthesis indicate the total for the system.
3. Simulation
In this section, Monte Carlo (MC) simulationsof the performance and characteristics of machined scintillators are discussed. These studies were per-formed using the Geant4 tool-kit, which simulatesthe passage of particles through matter [14]. Com-parisons are made with data observed in experi-ments conducted on the bench (Sec. 5.2) and withbeam data (Sec. 7).
As discussed in Sec. 2.2, the ST paddle geometryhas a nose section which tapers at the downstreamend. This causes the light collection efficiency ofhits in the nose section to increase as the hit posi-tion moves farther from the photo-detectors, con-trary to the usual behavior of scintillator material.A simple Geant4 simulation was conducted to in-vestigate the light collection efficiency. The detailsof the simulation are discussed in Ref. [2]. Only thetwo trapezoidal regions of a machined scintillatorpaddle were considered: the wide straight sectionand the tapered nose section.Ten thousand optical photons were generated at16 different locations inside the medium of the scin-tillator. The photon energies ranged between 0.5and 3.0 eV[15] and were generated isotropicallyfrom points along a 3 mm path in the scintilla-tor medium. This path is oriented orthogonal tothe wide surface of the scintillator. The number ofphotons detected by the SiPM, denoted as “SiPMHits”, is shown in Figures 8 and 9 as a functionof the source locations. For these studies, 100%detection efficiency was assumed for the simulatedSiPM. In the case of the nose section, the SiPM wasplaced at the wider upstream end of the simulatedscintillator bar. The results for this simulation arepresented in Fig. 8.The simulation shows that the tapering trape-zoidal geometry of the nose section results in im-proved light collection as the source moves furtheraway from the readout detector. There is an in-crease of 50% in light collection as the source ismoved from the near end to the far end of thenose section. The quasi-rectangular straight sectionshows the typical loss off light yield as the sourcemoves away from the photon detector.
Further simulations were conducted to study theeffects of the ST scintillator geometry and opti-cal surface quality on light collection. The scin-tillator geometry was imported into Geant4 from aVectorworks CAD drawing utilizing the CADMesh4
50 100 150 200 250 300 350 400
Source Distance (mm) S i P M H i t s Simplified Model
Nose SectionStraight Section
Figure 8: Simulation results for a simplified two section sce-nario. The total number of photons which were collectedby the SiPM detector for each of the 16 source locations isplotted against the source distance from the photon detector. utility[16]. The SiPM was modeled as a 12 × ×
10 mm volume with a 100 µ m air gap betweenit and the wide end of the straight section. Thevolume surrounding the scintillator was defined tobe air. The scintillator material, SiPM photon de-tector, and optical photons were defined in an amanner identical to that discussed in Sec. 3.1.To simulate the imperfections of scintillator sur-faces, an optical surface “skin” was defined. The“skin” conformed to the POLISH and UNIFIEDphysics models[17] and was of the type “dielectric-dielectric”. Both the transmission efficiency andreflection parameters were implemented as free pa-rameters in order to study their various effects onlight transmission.The POLISH model simulates a perfectly pol-ished surface while the UNIFIED model defines thefinish of the scintillator surface both of which areillustrated in Fig. 9 [17]. The details of the UNI-FIED model parameters are discussed in detail inReferences [2] & [17].As described in section 3.1, 10,000 optical pho-tons were generated in the scintillator medium ev-ery 2.5 cm and the number of hits in the SiPMwere recorded. For the POLISH model, only thetransmission efficiency (cid:15) was varied. For the UNI-FIED model, (cid:15) and the radiant intensity parameterswere held constant while σ α , which characterizesthe standard deviation of the surfaces micro-facetorientation, was varied. In both instances the at-tenuation length α was extracted in the straightsection. The results are shown in Fig. 10. Figure 9: POLISH and UNIFIED models of scintillator sur-faces. Left: Polar plot of the radiant intensity of the POL-ISH model. Right: Polar plot of the radiant intensity in theUNIFIED model [17]. (cid:126)d i , (cid:126)d r , (cid:126)d t are the incident, reflected,and refracted photon direction vectors respectively while (cid:126)σ i and (cid:126)σ r are the associated incident and reflected angles withrespect to the average normals. n and n are the indicesof refraction for the incident and transmission mediums re-spectively. R is the probability of Fresnel reflection at thesurface and the complementary probability of transmissionis simply T = 1 − R . For the POLISH model it is clear that if thetransmission efficiency increases, i.e. the reflectionefficiency decreases, the amount of light collected inthe SiPM decreases as illustrated in Fig. 10. Simi-larly, as the number of micro-facet orientations in-crease, meaning a more coarsely ground surface,the amount of light collection in the SiPM also de-creases. Moreover, good surface quality enhancesthe rise in light collection in the nose region.
4. Misalignment Studies
Here we discuss the relative alignment of a scin-tillator paddle with a SiPM detector and its effectson light collection and time resolution.
The SiPM was mounted atop a Newport MT-XYZ (MT) linear translation stage[18] with adjust-ment screws providing translations of 318 µ m per360 ◦ rotation. The SiPM collected light from a scin-tillator paddle at the upstream end of the straightsection. A Sr source and trigger photomultipliertube (PMT) were fixed 24 . µ m accuracy. Further detailsof the experimental set-up are discussed in Ref. [2].5
10 20 30 40 50 60 ± : 26.4 α : 0.100, α σ ± : 20.2 α : 0.075, α σ ± : 20.8 α : 0.100, α σ ± : 17.1 α : 0.125, α σ ± : 21.5 α : 0.150, α σ ± : 14.7 α : 0.175, α σ ± : 16.1 α : 0.200, α σ ) + C α f(x) = A exp(-x/ ε : 0.01, Finish: GroundSiPM: 12x12 mm, Lobe: 0.045, Spike: 0.95, Back: 0.005
0 10 20 30 40 50 60020040060080010001200 ± : 26.4 : 0.010, 6.2 cm ± : 28.1 : 0.020, 3.8 cm ± : 24.8 : 0.030, 2.0 cm ± : 18.2 : 0.040, 1.0 cm ± : 13.9 : 0.050, αε αε αε αε αε ) + C α f(x) = A exp(-x/SiPM: 12x12 mm, Finish: Polished S i P M H i t s Source Distance (cm) Source Distance (cm)
Figure 10: POLISH and UNIFIED model results. Shown is the number of hits recorded in the SiPM (vertical axis) versusthe source distance (x-axis). Left: POLISH model varying the transmission efficiency (cid:15) . Right: UNIFIED model varying thestandard deviation of the surfaces micro-facet orientation σ α .Figure 11: Optics setup for misalignment studies. Left:SiPM & scintillator vertical misalignment. Right: SiPM &scintillator horizontal misalignment. The scintillator remained fixed while the SiPMwas scanned across the upstream end of the scin-tillator (Fig. 11). During this scan, the horizontalalignment ( z ) of the SiPM and scintillator was fixedat a distance of 100 µ m and was monitored closely.At y = 0 the SiPM and scintillator are aligned verti-cally. The measurements and simulations are shownin Fig. 12. There is no significant variation of timeresolution within a ± µ m range of the optimalalignment.A Geant4 simulation, done in a manner similar tothat discussed in section 3.2, was utilized to studythe effect of vertical misalignment. The photon col-lection statistics at various y -positions in simula-tion matched data taken on the bench. Ergo, themeasured time resolution is dominated by photoncollection statistics. Thus, we determined the sim-ulated time resolutions empirically, by scaling lightcollection to the time resolutions measured on the Vertical Misalignment (mm) T i m e R e s o l u t i o n ( p s ) Vertical Misalignment Studies
Bench DataSimulated Data
Figure 12: Vertical misalignment results. The minimumtime resolution obtained was approximately 350 ps whichwas expected. Once the SiPM exceeded y = ± mm , noactive area of the SiPM was directly coupled to the face ofthe scintillator. bench. The acceptable range of vertical misalign-ment is approximately ± µ m. The effects of varying the horizontal alignmentwere also studied. While the horizontal alignment( z ) was varied, the vertical alignment ( y ) was keptconstant at the optimal location ( y = 0), and wasmonitored both optically and manually with a mi-crometer.6he SiPM was moved along the z -axis. We de-fined z = 0 to be the position where the active areaof the SiPM was flush against the face of the scin-tillator paddle. The results of this study are illus-trated in Fig. 13. While the simulation underesti- Horizontal Misalignment (mm) T i m e R e s o l u t i o n ( p s ) Horizontal Misalignment Studies
Bench DataSimulated Data
Figure 13: Horizontal misalignment results. mates the degradation of resolution with increasinghorizontal alignment, it is clear from the data thatthe optimal coupling range is z < µ m. More-over, there is no significant degradation in time res-olution for z < µ m.
5. Fabrication
The details of polishing and characterizing ma-chined scintillators, as well as the construction ofthe Start Counter are discussed.
While undergoing edge polishing at McNeil En-terprises, the machined scintillators incurred sur-face damage and were exposed to chemical contam-inants known to harm scintillator surfaces. Polish-ing was required to restore adequate performancecharacteristics.To polish the machined scintillator surfaces,Buehler Micropolish II deagglomerated 0 . µ m alu-mina suspension was utilized [19]. The polishingsuspension was diluted with a 5:1 ratio of de-ionizedH O to alumina and applied to a cold, wet 6 (cid:48)(cid:48) × . (cid:48)(cid:48) Caswell Canton flannel buffing wheel [20] operatedat speeds less than 1500 RPMs. The surfaces ofthe scintillators were carefully buffed until the largesurface defects were removed. In order to eliminatesmall localized surface defects, hand polishing with a soft NOVUS premium Polish Mate microfilamentcloth [21] and diluted polishing suspension was ap-plied. These polishing procedures made the scin-tillators void of most visible scratches and surfacedefects.The improved surface quality of the polished scin-tillators are shown in Fig. 14 where a scintillatorpaddle before and after polishing is shown. A red
Figure 14: Effects of polishing scintillators. Left: non-diffuselaser incident on an edge, before polishing, at the upstreamend of the straight section. Right: non-diffuse laser incidenton the same edge, after polishing, at the upstream end ofthe straight section. laser beam was shone into the scintillator mediumfrom the upstream end aimed at one edge. Theunpolished scintillator had such poor surface qual-ity that the reflections of the laser in the bend re-gion could not be resolved. However, the reflectionsin the polished scintillator can clearly be observedtraversing the bend and nose region. On average,at the tip of the nose, the scintillators exhibiteda 15% improvement in time resolution. Moreover,variation in performance from counter to counterwas substantially reduced.
The polished scintillators were tested for lightoutput and time resolution properties. A test stand(Fig. 15) was used to measure the response of ma-chined scintillators at four locations in the straightsection, three in the bend, and five in the nose.The measurements were conducted with a colli-mated Sr source oriented orthogonal to the wideflat surface of the scintillators. The Sr sourceprovides electrons ranging from 0 . − . igure 15: CAD Drawing of the scintillator test stand. light from the scintillator being tested. The ADCand TDC data were analyzed to determine the lightoutput and time resolution.The 30 machined scintillator paddles that exhib-ited the best time resolution and light output prop-erties from a set of 50 were selected for the final con-struction. These scintillators were then wrapped in16 . µ m thick reflective film (aluminum foil) andtested again. Their measured time resolutions areillustrated in Fig. 16. The phenomenon of increased Source Distance (cm) T i m e R e s o l u t i o n ( p s ) Nose RegionST Bench Performance σ avg = 359 ps Relative Spread of Time ResolutionsWeighted Avg. of Time Resolutions
Figure 16: Weighted average of the time resolution of 30scintillator paddles as a function of distance from the SiPM.The shaded vertical blue boxes indicate the relative spreadof the time resolutions among the 30 paddles. The dashedline indicates the weighted average over the 12 data points. light collection in the nose region is observed. Thelarger time resolution in the straight section is dueto light which initially travels downstream is re-flected from the nose.
To build the ST an assembly jig (Fig. 17) wasfabricated. The upstream support hub and Ro-
Figure 17: CAD drawing of the ST assembly jig. hacell support structure were attached to a rotat-ing bracket that moved in discretized 12 ◦ incre-ments. Two pneumatic cylinders with soft, semi-dense rubber feet were used to hold a single scintil-lator in place. Two free floating acrylic rings, with30 tapped holes 12 ◦ apart, housed 10 ◦ swivel padthumb screws fitted with silicone foam. The thumbscrews held installed paddles in place.A camera was used to measure and controlthe scintillator/SiPM vertical and horizontal align-ments. Vertical alignment was achieved by usingKapton shims between the scintillator and the sup-port structure. The horizontal alignment was con-figured to a distance less than 200 µ m between thescintillator and the SiPMs.To secure paddles the to the Rohacell supportstructure the ST was wrapped around its circumfer-ence using self-adhesive transparent bundling wrap(0.8 mil thick, 6 in wide) at six locations alongthe length of the detector as seen in Fig. 18. De- Figure 18: ST assembly before (A) and after (B) wrappingwith bundling wrap
GlueX liquid H target as shown in Fig. 19. Figure 19: ST mounted the
GlueX target. The beam direc-tion is from right to left and travels down the central axisof the ST. During operation the ST resides in the bore ofthe central tracking chamber, which is visible in the top leftcorner.
6. Calibration
The procedures to optimize the time resolutionfor particle identification (PID) and time of flight(TOF) are discussed here.
To correct the TDC timing for variations dueto pulse shape, we use the timing signal from theFADC250s. The latter uses a digital algorithmsimilar to a constant-fraction discriminator andtherefore gives a time largely independent of pulseheight[2][24]. The TDC/FADC time difference isgiven by Eq. 1 where i is the paddle number index. δt i = t TDC i − t FADC i (1)The FADC250’s report the amplitude, integral,and time of the input analog signals[24]. The am-plitude was selected for the time-walk correctionsbecause it is correlated better with the leading edgetime of the pulse[2]. Figure 20 (left) shows a typicaltime-walk spectrum, i.e. δt versus the pulse ampli-tude, for one paddle of the ST. This correlation canbe described empirically by the function given byEq. 2 [25] where a and a thresh i are the pulse ampli-tude and discriminator threshold respectively, and c i , c i , c i are the fit parameters. f wi (cid:0) a/a thresh i (cid:1) = c i + c i ( a/a thresh i ) c i (2) The most probable value (MPV) of the pulse am-plitude spectra was chosen as the reference pointwhere the time-walk correction is defined to be zero.Fig. 20 (right) illustrates the effect on the time dif-ference spectrum ( δt ) as a result of the applied time-walk corrections. The time between the production of scintillationlight in a ST scintillator paddle and detection by theSiPM depends on the hit location along the paddleand is discussed below.The EJ-200 scintillator material has a refractiveindex of 1.58 [5] and the corresponding speed oflight in that medium is 19 cm / ns. The observedeffective velocity is slower. Correcting for this lightpropagation in the scintillator is necessary since theST paddles are 60 cm long. Studies showed thatthe effective velocity of light depends on the regionalong the paddle where the hit occurred. The prop-agation time corrections were conducted with well-defined reconstructed charge particle tracks. Fur-ther details regarding the event and track selectionare found in Ref. [2].The propagation time T STprop is determined byEq. 3 where T SThit is the time-walk corrected hit time, T STflight is the flight time from the track vertex to theST intersection point, and T BBvertex is the track vertextime. T STprop = T SThit − T STflight − T BBvertex (3)The z -coordinate of the track’s intersection pointwith the ST ( z SThit ) are determined by the detectorgeometry as well as the distance d SThit of this inter-section point and the SiPM.The propagation times were determined in threedistinct regions corresponding to the three geomet-rical sections of the ST: the straight, bend, andnose regions. The propagation times in these re-gions were fit with a linear function given by Eq. 4where j indicates which region in the i th paddle isbeing fit and A and B are fit parameters. f ij ( z ) = A ij + B ij · z (4)Figure 21 (left) illustrates the correlation betweenthe propagation time and the distance from theSiPM with T STprop = 0 . d SThit = 0 . .5 1.0 1.5Pulse Amplitude (V)2.01.51.00.50.00.51.01.52.0 t ( n s ) Before Time Walk Correction t ( n s ) After Time Walk Correction Figure 20: Left: Single paddle time-walk spectrum; the line shown is the fitted function used to determine the correctionfactors. Right: after time-walk correction. Plotted on the vertical axis is δt and on the horizontal axis is the correspondingpedestal subtracted pulse amplitude spectrum. T i m e ( n s ) Before Propagation Time Correction
Straight SectionBend SectionNose Section T i m e ( n s ) After Propagation Time Correction Figure 21: Left: Single paddle propagation time correlation. T STprop is plotted on the vertical axis and d SThit is plotted along thehorizontal axis. There is a clear correlation between the time when optical photons are detected by the SiPM and the locationof the scintillation light along the length of the paddle. Right: Single paddle propagation time after correction.
To measure attenuation in the scintillators,charged tracks were selected in a manner similarto that discussed in Sec. 6.2. The uncorrected en-ergy deposition ( dE M ) per unit length ( dx ) versusthe track momentum ( p ) for tracks intersecting tothe ST are shown in Fig. 25. It is clear that no re-liable PID can occur for tracks with p > . / cwithout further corrections.The pulse integral (PI) data, normalized to thepath length dx of the track in the scintillatormedium, were binned in 3.5 cm z SThit bins along thelength of the paddle. The MPV of the PI was extracted utilizing an empirical function given byEq. 5 where p , p , p are the fit parameters. f ( z ) = p e ( − p ( z − p )) × (1 + tanh( p ( z − p ))) (5)A fit to the data in a single 3.5 cm z SThit bin is illus-trated in Fig. 22. The MPV was extracted analyti-cally and then plotted against the average value foreach z SThit bin as shown in Fig. 23.In order to characterize the photon attenuation,the straight and nose regions were treated indepen-dently. The piecewise continuous function given byEq. 6 was selected to fit the data where the in-tersection Z ib (or correction boundary) of the two10
000 5000 6000 7000 8000 9000dE/dx (au)20040060080010001200 C o un t s MPV Extraction
Figure 22: Pulse integral integral data normalized to the thetrack length in the scintillator medium for a single 3.5 cmbin along the paddle length.
20 25 30 35 40 45 50 55 60Source Distance (cm)400050006000700080009000100001100012000 d E / d x ( a u ) Attenuation Correction
Straight section fit Nose section fitBefore CorrectionAfter Correction
Figure 23: Fits to the attenuation data. exponential fit functions was fixed and is shown inFig. 23. f ic ( z ) = (cid:40) A iS e − B iS · z z ≤ Z ib cm A iN e B iN · z + C iN z > Z ib cm (6)In Eq. 6, the subscripts S and N denote the straightand nose sections respectively while A i , B i , and C i are the fit parameters for the i th paddle.An attenuation correction factor R i ( z ) is ap-plied to the deposited energy measurement per unittrack-length ( dE M /dx ) to give the corrected energydeposition per unit track length ( dE iC ( z ) /dx ) forpaddle i and is given by Eq. 7 where the subscripts C and M are the corrected and measured quantitiesrespectively. dE iC ( z ) dx = dE M dx · R i ( z ) = dE M dx · f ic (0) f ic ( z ) (7)After attenuation corrections are applied, particle separation is greatly improved. This will be dis-cussed further in Sec. 7.
7. Performance
The increase in light output as a function ofhit position along the ST detector during nominal
GlueX beam conditions is illustrated in Fig. 24.This is advantageous because the majority of the P u l s e A m p li t u d e ( V ) Light Collection Figure 24: Typical FADC250 pulse amplitude spectrum ver-sus the z -component of charged tracks intersecting the STfor an individual ST sector. The vertical line indicates thestart of the tapered nose section. charged tracks produced intersect the ST in the for-ward region.With the attenuation corrections discussed inSec. 6.3 applied to the data, the PID capabilitiesof the ST were improved. Figure 25 illustrates thePID capability of charged tracks intersecting theST. Protons can be separated from other hadronswith momenta up to 0.9 GeV/c which is a factor1.5 improvement relative to the uncalibrated data.The PID capabilities of the ST extend the identifi-cation of low momentum protons that do propagatethrough the central drift chambers.The ST was used to determine the time of the in-teraction of a beam photon with the LH target af-ter the time-walk and propagation time correctionsdiscussed in Sec. 6.1 and 6.2 were applied. Theinteraction time can be determined independentlyfrom a timing signal originating from the accelera-tors RF system, the latter with very high precision.The time difference between the ST time and themachine RF time is shown in Fig. 26. The acceler-ator can be run in mode where the time separationbetween beam bunches is 2 ns, a separation indi-cated in the figure. One application of the ST is to11 .00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00Momentum (GeV)0100002000030000400005000060000 d E / d x ( a u ) Particle Identification d E / d x ( a u ) Particle Identification Figure 25: Left: Typical uncorrected dE/dx vs. p distribution. Right: Corrected dE/dx vs. p distribution. The “bananaband” corresponds to protons while the horizontal band corresponds to electrons, pions, and kaons. It is clear that after thecorrections have been applied, pion/proton separation is achievable for tracks with p < . / c. distinguish particles from different RF buckets onthe basis of timing. N o r m a li z e d C o un t s Time Resolution
Figure 26: Time resolution for one paddle with its full widthhalf maximum value indicated in ns. The x-axis is the timedifference between T STvertex and T BBvertex . The vertical linesindicate the cuts used to identify a 500 MHz beam bunch.
The measured distribution is fit with the sum oftwo Gaussians and the full width at half maximum(FWHM) of the resulting curve is calculated. Also,the fraction of the area of the curve within ± Section All Straight Bend NoseFWHM
550 ps 690 ps 700 ps 450 ps
Fraction
93% 92% 91% 94%
Table 2: Average time resolutions (FWHM) and event frac-tions within a ± sections combined is due to the majority of eventsintersecting the ST in the nose section. It is clearfrom Table 2 that measurements made with beamdata also exhibit improved light collection, and thusimproved time resolution, in the nose region.Approximately four years of operation haveelapsed since the paddles were first tested on thebench at Florida International University (FIU).Prior experience with scintillators indicates thatdegradation in time resolution as a result of mis-handling will be visible in a matter of weeks. Nodegradation in time resolution has been observedand the ST is still performing well below designresolution.
8. Conclusion
The
GlueX
Start Counter was designed and con-structed at Florida International University for usein Hall D at TJNAF. It provides separation of the500 MHz photon beam bunch structure delivered bythe CEBAF to within 94% accuracy. It is the first“start counter” detector to utilize magnetic field12nsensitive SiPMs as the readout system. Despitethe many design and manufacturing complications,the ST has proven to have performed well beyondthe design resolution of 825 ps (FWHM) with anaverage measured resolution of 550 ps (FWHM).Furthermore, the capabilities of the ST make it aviable candidate to assist in particle identification.The unique geometry of the ST nose section hasillustrated the advantage of tapering trapezoidalgeometry in thin scintillators. Through simula-tion, tests on the bench, and analysis of data ob-tained with beam, it has been definitively demon-strated that this geometry results in a phenomenonin which the amount of light detected increases asthe scintillation source moves further downstreamfrom the readout detector.Since its installation in Hall D during the Fall2014 commissioning run, the ST has shown no mea-surable signs of deterioration in performance. Thissuggests that the ST scintillators are void of crazingand will most likely be able to meet and exceed thedesign performance well beyond the scheduled runperiods associated with the
GlueX experiment.It is planned to incorporate the ST into the level1 trigger of the
GlueX experiment for high lumi-nosity running when there will be 5 × γ/s in thecoherent peak. Preliminary studies suggest thatwhile operating at rates in excess of 300 kHz perpaddle, the ST exhibits a high efficiency ( > GlueX .
9. Acknowledgments
The authors would like to graciously thank theplethora of Jefferson Lab staff members in both theEngineering division and Hall-D. Their numerousand invaluable contributions allowed for the StartCounter project to come to fruition. The authorswould also like to extend their gratitude to theentire
GlueX
Collaboration who provided fruit-ful ideas and advice throughout the many stagesof the project. Work at Florida International Uni-versity was supported in part by the Departmentof Energy under contracts DE-FG02-99ER41065 and DE-SC00-13620. Furthermore, this materialis based upon work supported by the U.S. De-partment of Energy, Office of Science, Office ofNuclear Physics under Contract No. DE-AC05-06OR23177.13 eferences [1] H. Al Ghoul, et al., First results from the gluex ex-periment, AIP Conference Proceedings 1735 (1) (2016)020001. arXiv:https://aip.scitation.org/doi/pdf/10.1063/1.4949369 , doi:10.1063/1.4949369 .URL https://aip.scitation.org/doi/abs/10.1063/1.4949369 [2] E. Pooser, The GlueX start counter and beamasymmetry Σ in single π photoproduction, Ph.D.thesis, Florida International University, http://digitalcommons.fiu.edu/etd/2450/ (2016).[3] A. Somov, Performance of the level-1 trigger athigh luminosity, talk Given at the January 19,2017 GlueX Collaboration meeting available at https://halldweb.jlab.org/DocDB/0031/003193/004/high_rate_status.pdf (2017).[4] .[5] General Purpose Plastic Scintillator EJ-200, EJ-204, EJ-208, EJ-212, technical note available at .[6] .[7] .[8] .[9] .[10] F. Barbosa, Time characteristics of silicon photomul-tipliers used in the gluex experiment, in: 2015 IEEENuclear Science Symposium and Medical Imaging Con-ference (NSS/MIC), 2015, pp. 1–4, available at http://ieeexplore.ieee.org/document/7581740/ .[11] Multi-Pixel Photon Counter, technical note avail-able at https://halldweb1.jlab.org/wiki/images/4/49/S10362-33_series_kapd1023e05.pdf .[12] F. B. et al., A VME64x, 16 Channel, Pipelined 250MSPS Flash ADC With Switched Serial (VXS) Ex-tension, Jefferson Lab, technical note GlueX-doc-1022available at https://halldweb.jlab.org/doc-public/DocDB/ShowDocument?docid=1022 (April 2008).[13] F. B. et al., The Jefferson Lab High Resolution Time-to-Digital Converter (TDC), Jefferson Lab, technical noteGlueX-doc-1021 available at https://halldweb.jlab.org/doc-public/DocDB/ShowDocument?docid=1021 (April 2008).[14] https://geant4.web.cern.ch/geant4/ .[15] K. S. Krane, Introductory Nuclear Physics, John Wiley& Sons, 1988, Ch. Detecting Nuclear Radiations, pp.201–202.[16] C. M. Poole, A cad interface for geant4, paper submit-ted to Cornell Archive (arXiv) May 5, 2011 available at http://arxiv.org/pdf/1105.0963.pdf (2011).[17] A. L. . C. Moison, A more physical approach tomodel the surface treatment of scintillation countersand its implementation into detect, paper presented atIEEE, Nuclear Science Symposium; Anaheim, Novem-ber 2-4 available at http://geant4.slac.stanford.edu/UsersWorkshop/PDF/Peter/moisan.pdf (1996).[18] .[19] https://shop.buehler.com/consumables/grinding-polishing/polishing-suspensions/alumina-suspensions .[20] . [21] .[22] Sr Decay Radiation Data, national Nuclear DataCenter, Nuclear Structure and Decay Database(NuDat) available at .[23] Y Decay Radiation Data, national Nuclear DataCenter, Nuclear Structure and Decay Database(NuDat) available at .[24] H. Dong, Description and Instructions for theFirmware of Processing FPGA of the ADC250 BoardsVersion 0x0C0D, fADC250 Manual Available at https://coda.jlab.org/drupal/system/files/pdfs/HardwareManual/fADC250/FADC250_Processing_FPGA_Firmware_ver_0x0C0D_Description_Instructions.pdf (February 2017).[25] E. Smith, Low-level calibration constants for BCAL,technical note GlueX-doc-2618-v10, available at https://halldweb.jlab.org/DocDB/0026/002618/011/bcal_constants.pdf (December 2014).(December 2014).