Design and characterization of a single photoelectron calibration system for the NectarCAM camera of the medium-sized telescopes of the Cherenkov Telescope Array
Barbara Biasuzzi, Kevin Pressard, Jonathan Biteau, Brice Geoffroy, Carlos Domingues Goncalves, Giulia Hull, Miktat Imre, Michael Josselin, Alain Maroni, Bernard Mathon, Lucien Seminor, Tiina Suomijarvi, Thi Nguyen Trung, Laurent Vatrinet, Patrick Brun, Sami Caroff, Stephen Fegan, Oscar Ferreira, Pierre Jean, Sonia Karkar, Jean-François Olive, Stéphane Rivoire, Patrick Sizun, Floris Thiant, Adellain Tsiahina, François Toussenel, Georges Vasileiadis
DDesign and characterization of a single photoelectron calibration systemfor the NectarCAM camera of the medium-sized telescopes of theCherenkov Telescope Array
Barbara Biasuzzi a , Kevin Pressard a , Jonathan Biteau a , Brice Geo ff roy a , Carlos Domingues Goncalves a ,Giulia Hull a , Miktat Imre a , Michael Josselin a , Alain Maroni a , Bernard Mathon a , Lucien Seminor a , TiinaSuomijarvi a , Thi Nguyen Trung a , Laurent Vatrinet a , Patrick Brun e , Sami Caro ff b,1 , Stephen Fegan b , OscarFerreira b , Pierre Jean f , Sonia Karkar c , Jean-Franc¸ois Olive f , St´ephane Rivoire e , Patrick Sizun d , FlorisThiant b , Adellain Tsiahina f , Franc¸ois Toussenel c , Georges Vasileiadis e a Institut de Physique Nucl´eaire, IN2P3 / CNRS, Universit´e Paris-Sud, Universit´e Paris-Saclay, 15 rue Georges Cl´emenceau, 91406Orsay, Cedex, France b Laboratoire Leprince-Ringuet, ´Ecole Polytechnique (UMR 7638, CNRS / IN2P3, Universit´e Paris-Saclay), 91128 Palaiseau, France c Sorbonne Universit´es, UPMC Universit´e Paris 06, Universit´e Paris Diderot, Sorbonne Paris Cit´e, CNRS, Laboratoire de PhysiqueNucl´eaire et de Hautes Energies (LPNHE), 4 Place Jussieu, 75252, Paris Cedex 5, France d IRFU, CEA, Universit´e Paris-Saclay, 91191 Gif-sur-Yvette, France e Laboratoire Univers et Particules de Montpellier, Universit´e de Montpellier, CNRS / IN2P3, CC 72,Place Eug`ene Bataillon, F-34095Montpellier Cedex 5, France f Institut de Recherche en Astrophysique et Plan´etologie, CNRS-INSU, Universit´e Paul Sabatier, 9 avenue Colonel Roche, BP 44346,31028 Toulouse Cedex 4, France
Abstract
In this work, we describe the optical properties of the single photoelectron (SPE) calibration systemdesigned for NectarCAM, a camera proposed for the Medium Sized Telescopes (MST) of the CherenkovTelescope Array (CTA). One of the goals of the SPE system, as integral part of the NectarCAM camera,consists in measuring with high accuracy the gain of its photo-detection chain. The SPE system is basedon a white painted screen where light pulses are injected through a fishtail light guide from a dedicatedflasher. The screen – placed 15 mm away from the focal plane – is mounted on an XY motorization thatallows movements over all the camera plane. This allows in-situ measurements of the SPE spectra via acomplete scan of the 1855 photo-multiplier tubes (PMTs) of NectarCAM. This calibration process willenable the reduction of the systematic uncertainties on the energy reconstruction of γ -rays coming fromdistant astronomical sources and detected by CTA.We discuss the design of the screen used in the calibration system and we present its optical performancesin terms of light homogeneity and timing of the signal. Keywords:
CTA, NectarCAM, Medium Sized Telescope
1. Introduction
The current most sensitive γ -ray telescopes em-ploy the so-called imaging atmospheric Cherenkovtechnique that allows γ -rays to be indirectly de-tected through the light yield of their atmosphericshowers [1]. A γ -ray entering the atmosphere in-teracts with the Coulomb field of nuclei in themedium, and generates a shower of secondary par-ticles. The velocity of e − and e + in the shower canbe higher than the speed of light in the air, so they Currently a ffi liated to c . induce the emission of a Cherenkov-light cone thatpropagates towards the ground and can be observedwith dedicated telescopes. The currently operat-ing ground-based γ -ray telescopes are VERITAS, H.E.S.S., MAGIC, FACT, and the future gener-ation is represented by the Cherenkov TelescopeArray (CTA) [2, 3], which will enable observa-tions in the interval between 30 GeV and 300 TeV https://veritas.sao.arizona.edu/ https://magic.mpp.mpg.de/ Accepted for publication in NIMA on 9th October 2019 doi: 10.1016 / j.nima.2019.162949 a r X i v : . [ a s t r o - ph . I M ] O c t (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ (a) (b)
190 mm mm (c) Figure 1: (a) bottom view of the assembly of the flasher box containing the LED electronic card, the fishtail light guide, and the finalversion of the screen; (b) front view of the camera focal plane with the SPE calibration system in parking position on the bottom-rightcorner (di ff erent colors are used to distinguish the PMT modules); (c) simplified lateral view of the front of the camera, showing thescreen in a possible position during a camera scan: the emitting side is facing the PMTs, while the reflective one is looking at themirror dish of the telescope. Two distances are reported: the distance between the entrance of the Winston cone and the edge of thescreen facing the mirrors (25.6 mm), and the minimum distance between the camera window (gray band) and the screen (190 mm). with a sensitivity improved by a factor 5-20 de-pending on energy. CTA will consist of two arrayswith telescopes of three di ff erent sizes: the LargeSized Telescopes (LST) with a 23 m-diameter, theMedium Sized Telescopes (MST) with a 12 m-diameter, and the Small Sized Telescopes (SST)with a 4 m-diameter. In order to cover the wholesky, the two arrays will be placed in the Northernhemisphere in La Palma (Canary Islands, Spain)and in the Southern hemisphere, at Paranal (Chile),respectively.The reconstruction of the γ -ray properties, suchas arrival direction and energy, is not trivial, sincethe Cherenkov telescopes rely on an indirect detec-tion technique. An important source of systematicuncertainty derives from the knowledge of the gainof the whole photo-detection chain, including de-tectors such as the photo-multiplier tubes (PMTs),which detect the Cherenkov light produced by theparticle showers.The calibration process is crucial to improvethe performances of current-generation Cherenkovtelescopes and of the future CTA in the reconstruc-tion of the γ -ray properties. In particular, varioustechniques are employed to monitor the gain of thePMTs.VERITAS relies on a LED-based flasher sys-tem to measure the gain of the PMTs throughboth the photostatistics method and the single-photoelectron (SPE) spectrum fitting [4]. In thefirst method, the gain is estimated by exploiting therelation between the mean number of photoelec-trons hitting the first dynode and the width of themeasured charge distribution [5], [6]. In the secondmethod, SPE measurements are performed in-situby placing a thin aluminum plate with 3 mm holes aligned with the center of each PMT. The gain de-termination through the SPE fitting is also adoptedby H.E.S.S. [7, 8]. The system used at H.E.S.S.consists of a pulsing LED and a di ff user to homo-geneously illuminate the camera, placed in a shel-ter 2 m away from it. Finally, MAGIC uses stronglight pulses, with three methods for the absolutelight flux calibration [9, 10, 11]. In the first method,the intensity of the pulse is obtained from the SPEspectrum of a so-called “blind pixel”, darkened bya calibrated filter, while in the second one it is re-trieved through a comparison with a calibrated PINdiode. Once the light intensity is known, the gain ofall the PMTs can be retrieved. In the third method,the one currently in use, the gain of the PMTs canbe determined through the photostatistics method,as done by VERITAS.The multi-component nature and the large flex-ibility in operation of CTA will require di ff erentcalibration systems, techniques, and strategies forthe camera calibration, based on method of pho-tostatistics, single photoelectron acquisitions, andobservational data such as muon rings [12, 13, 14].For the MSTs of CTA, we developed an SPE cal-ibration system that enables the measurement ofthe gain of the whole photo-detection chain witha statistical uncertainty lower than 1%. The SPEcalibration system is integrated in the NectarCAMproject [15], one of the cameras proposed for theMSTs [3]. NectarCAM has been designed tocover the energy range between 100 GeV and 30TeV, with a wide field of view of 8 ◦ . Nectar-CAM encompasses 265 di ff erent modules: each The other camera proposed for the modified Davies-CottonMSTs is the FlashCam [16]. (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ of them includes PMTs, HV dividers, high voltagesupply, pre-amplifier, trigger, readout, and Ether-net transceiver. Each module hosts 7 PMTs fromHamamatsu (R12992-100), with 7 dynodes each,for a total of 1855 PMTs in the whole camera. Infront of each PMT, a Winston cone is mounted inorder to maximize the light collected at the photo-cathode of each PMT.
2. Overview of the SPE calibration system
Both for the mirror alignment and the pointspread function (PSF) studies, a movable systemis required on the focal plane. The study of thegain of the PMTs can be done exploiting the samesystem. This brings another important advantage:the gain estimation can be performed with the cam-era closed (including during daytime, provided alightproof camera enclosure), without contamina-tion from the night sky background. The dual pur-pose of the SPE calibration is accomplished bythe two sides of its screen: the side directed to-wards the PMTs will enable the measurement ofthe gain of the whole photo-detection chain in SPEregime, while the other side will be used for mirroralignment and the study of the optical PSF. Thedeveloped system is formed by two main compo-nents: (i) a source light box that produces flashesof light, at a typical wavelength of 390 nm, whichare injected into (ii) a 10 mm thick screen made ofPoly(methyl methacrylate) (PMMA) covering anarea equivalent to 51 PMTs, or ∼ (Fig. 1a).The system is mounted inside the camera (15 mmaway from the Winston cones) on a system of xy rails, and moved by motors over the entire focalplane. When not in use, it is located in a parkingposition, on the bottom-right corner of the camera,as shown in Fig. 1b. Figure 1c shows the positionof the screen in a lateral view of the camera.Both sides of the screen are painted with a highlyreflective paint. The side facing the focal plane ispainted with a dedicated pattern to optimize the ho-mogeneity of light emission, while the side facingthe mirrors is covered with three layers of paint.Both the mirror alignment and the PSF stud-ies can be performed by analyzing the image ofa point-like light source that forms on the side ofthe screen facing the mirrors. A movable system The optical PSF refers to the instrumental response to a dis-tant point-like source emitting in the visible spectrum, and forCherenkov telescopes it is of the order of 1 arcmin. The γ -PSFrefers to the accuracy in the reconstruction of the γ -ray direc-tion, and it is of the order of 0.05-0.1 deg. enables the study of the PSF both on- and o ff -axis. The point-like source image is captured bya charge-coupled device camera located in the cen-ter of the mirror dish, and further analyzed. Themirror alignment can be performed by using dif-ferent methods, such as the 2f alignment [17], theSCCAN method [18] that tracks a star and observesthe reflections of the mirrors on a camera placed inthe reflectors focal point, or the Bokeh method [19]that analyzes the blurring of an image on the focalplane. Finally, as in the MST case, the alignmentmethods use remotely controllable orientation ac-tuators on the mirrors in order to optimize the liveimage of a star on the dedicated screen [20].The emission towards the focal plane, neededfor the gain estimation, is obtained through the di-rect injection of light into the screen. For this aim,a dedicated light box source has been designed atthe Laboratoire Univers et Particules de Montpel-lier (LUPM). It is equipped with 12 light emittingdiodes (LEDs) at 390 nm pulsing with adjustablefrequency and amplitude. The typical full width athalf maximum (FWHM) of the pulses is between 3and 5 ns. The light produced by the flasher is con-veyed and injected into the screen through a fishtaillight guide, made of the same PMMA material asthe screen. To save as much space as possible infavor of a larger screen-surface, the width and thelength of the fishtail light guide are set to equal val-ues of 148.1 mm. The input diameter of the fishtailis equal to 50 mm.To achieve a low-intensity emission towards thefocal plane, the screen is entirely covered with areflective paint that forces the light to be reflectedseveral times inside the screen, so that only a smallfraction can escape from its surface. With this de-sign, low-intensity flashes illuminate the PMTs sothat the gain can be measured acquiring SPE spec-tra. To facilitate the light injection, the fishtaillight guide is painted with the same material as thescreen, and then wrapped with a black tape to avoidlight leakage.
3. Requirements for the di ff use-reflective screen In the conceptual design of the di ff usive-reflective screen, we needed to account for threemain requirements from the project: Standard 3 mm UV LEDs (ref. 3RS4VCS): wavelength390 nm, typical forward voltage 3-4 V, opening angle 30 ◦ , lightintensity 0-200 mCd, driven voltage 7.5-16 V. (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ - the dimension of the screen must be greaterthan a disk with a diameter of 170 mm (cor-responding to 0.6 deg projected on the sky)to perform correctly the mirror alignment andthe study of the optical PSF;- for the PSF studies, the screen surface has toprovide a Lambertian di ff usive reflectivity >
90% between 450 and 700 nm.- the distance between the focal plane and thereflective side of the screen (including thethickness of the screen) must be 25.6 mm.This extra distance is needed to place thescreen in the image plane of distant stars,since the front of the Winston cone is fo-cused at ∼
10 km, the typical altitude of themaximum development of atmospheric air-showers.We aimed to build a screen with as homogeneousan emission as possible to enable SPE acquisitionover all its surface. We investigated systemati-cally the light-intensity contour maps to evaluatethe amount of area whose emission is within a cer-tain fraction of the maximum value measured onthe screen.Starting from these criteria, the design processfocused on the following aspects:- geometry: whatever the shape of the screen, itshould be significantly larger than the PSF. Itshould be as large as possible to enable a fastscan of the camera, and it must not shadow thefocal plane once in parking position.- coating type: it should be characterized by ahigh reflectivity both to ensure a low-intensityemission towards the focal plane, and to per-form the mirror alignment on the other side;- coating pattern: a study of the coating pat-tern has been performed in order to achievean emission as homogeneous as possible;- coating application: di ff erent types of coating-application methods have been investigated,not only to improve the light homogeneity, butalso to ensure high reproducibility and stableoptical performances.All these aspects have not been investigated sepa-rately, but they entered in di ff erent moments of theR&D phase of the reflective screen design, as ex-plained in the following sections.
4. Experimental setup
In the R&D phase of the SPE calibration sys-tem, we tested several screen prototypes, di ff erent Figure 2: Test-bench equipment placed in the dark room. Asystem of two motors moves a NectarCAM front-end board,equipped with only one PMT over the screen. Light pulses areinjected from the flasher into the fishtail light guide (black piece)and, in turn, into the screen. The LED electronic card of theflasher is stored inside the aluminum box attached to the fish-tail. in shape, painting-type coverage, painting pattern,and painting application method. In particular, asdescribed in the following sections, we investigatedfour geometries (square, rectangular, circular, andoctagonal), two reflective materials (polytetraflu-oroethylene and reflective painting), three coatingmethods (brush, air brush, and dip-coating), anddedicated painting patterns for di ff erent prototypes.For the tests, the screens were all placed in adark room. A NectarCAM front-end board [21](FEB, originally developed by a consortium involv-ing Institut de recherche sur les lois fondamentalesde l’Univers – IRFU –,
Laboratoire de physiquenucl´eaire et de hautes ´energies – LPNHE –,
In-stituto de Ciencias del Cosmos, Universitat deBarcelona – ICCUB –, and LUPM), is equippedwith one PMT (Hamamatsu R12992-100) placed ata fixed distance of 15 mm from the screen surface.They were moved by motors (drylin R (cid:13) stepper motorNEMA 23XL) on a 60 cm ×
40 cm system of rails(model ZLW-1040-S with 100 mm-long trolley) allaround the screen to acquire light intensity mea-surements. Motors are controlled via a dedicatedsoftware and guarantee a positioning accuracy of ± .
035 mm. A picture of the test bench is shownin Fig. 2.In order to ensure a positioning reproducibility dur-ing the test bench activities, a dedicated mask wasbuilt for each prototype. A grid drawn on suchmasks provided a way to cross-reference the mo-tors and the screen frame. We estimate the accuracyof the screen positioning of ± x and y axes.4 (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ The light pulses injected into the screen are pro-duced by a flasher developed at LUPM. The light isproduced by 12 LEDs emitting at 390 nm, the typ-ical wavelength of the Cherenkov light. The inten-sity can be tuned changing the operating voltage ofthe LEDs between 7.5 and 16.5 V. An extra LED in DC mode at longer wavelength ( ∼
660 nm) hasbeen added to simulate the night sky background[22]. The intensity of the light pulses (both in termsof number of LEDs, and their voltage) can be ad-justed through a labview software interface or anOPC UA server. The typical pulse frequency wasset to 100 Hz, and the trigger pulse duration to 250ns.Several versions of the flasher have been produced.Depending on the version of the test-bench setup,di ff erent neutral density filters (THORLABS filterswith optical density, OD, 0.5, 1.0, 2.0, 3.0, respec-tively) have been used to achieve the optimal emis-sion. The final version of the flasher includes a fil-ter with OD = ∼ ∼
50 photoelectrons (12LEDs at maximum voltage). By removing the fil-ter, the intensity can be increased up to ∼
500 pho-toelectrons.For each set of measurements, we adjusted theintensity of the light pulses depending on the PMTvoltage (typically between 1000 and 1400 V) withan appropriate intensity of the LEDs. The nominalhigh-voltage for the NectarCAM PMTs in operat-ing conditions is around 1000 V. With an appropri-ate intensity of the LEDs, we could acquire spectraboth in the faintest and in the brightest regions ofthe screen.Light intensity measurements were performed inhigh intensity regime, typically >
100 photoelec-trons produced at the photocatode. Measurementswere taken on a grid of points (usually 4 cm × Kingbright LED (ref. 934LSRD): wavelength 660 nm, typ-ical forward voltage 1.85 V, opening angle 60 ◦ , light intensity 0-20 mCd at 2 mA, driven in continuous current mode 0-12.5 mA. [ns] Figure 3: Example of PMT traces induced by the flasher. Thetraces were collected in a single position over the screen, andthe signal was integrated over a 60 ns window. shows the traces collected over a single position ofthe screen.
5. Results
At first, a small 20 cm ×
20 cm reflective screenprototype was built to validate the concept de-sign. Such a surface enables the calibration ofone entire module of the camera, or 11 individ-ual PMTs. Di ff erent di ff usive materials have beentested with this first prototype. In particular, weused a ∼ µ m polytetrafluoroethylene foil and aBicron paint (Saint Gobain BC-620) applied with abrush. This paint was selected for its high Lamber-tian di ff usive reflectivity ( >
90% above 400 nm)[23], its cost, and its ease of use. We also usedthis paint to paint the fishtail light guide; it wasthen covered with black tape to avoid light leak-age. Furthermore, several combinations of top-bottom and edge coverage of the screen have beenexplored:1. polytetrafluoroethylene on the bottom, blacktape on the edges (PTFE + BT);2. paint on the bottom, black tape on the edges(PAINT + BT);3. paint + aluminum foil (PAINT + AF);4. paint on the bottom, paint on the edges(PAINT + PAINT);5. paint on the bottom, polytetrafluoroethyleneon the top, paint on the edges (PAINT + PAINT + PTFE);6. paint on the bottom, paint on the top, paint onthe edges (PAINT + PAINT + PAINT). The performances of the black tape as a function of the ag-ing and di ff erent environmental conditions will be explored forin-situ operations and could be replaced by black paint if needbe. (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ Longitudinal distance [cm] I n t en s i t y [ A DC ]
1) PTFE+BT2) PAINT+BT3) PAINT+AF4) PAINT+PAINT5) PAINT+PAINT+PTFE6) PAINT+PAINT+PAINT
Painting (bottom-, edge-, up-side)
Figure 4: Light-intensity profiles measured along the centralaxis of the square screen. Measurements were taken at a dis-tance of 5, 10, and 15 cm from the injection edge, and are ex-pressed in analog-to-digital counts (ADC). Measurements weretaken switching on 1 LED at 8 V, with a filter of OD = Figure 4 shows the light-intensity profiles mea-sured along the central axis, perpendicular to thelight injection edge, for the above-mentioned con-figurations. For each configuration, the attenuationlength was estimated through a linear regression,and results are reported in Table 1. The estimateduncertainty on the slope is smaller than 0.02 cm forall measurements.Among the first four configurations that coveronly the bottom and the edges, those involvingPTFE + BT and PAINT + PAINT are able to min-imize the emission decrease with increasing longi-tudinal distance. To better control the emission (inorder to reach the SPE regime), and ensure that thelight would be carried to longer distances in largerscreens, we observed that a coverage on the top sidewas also necessary (configurations 5 and 6). Bothconfigurations provide a similar attenuation length,and capability to spread the light. However, afterthe first uses of these materials it became evidentthat it was quite di ffi cult to properly glue a polyte-trafluoroethylene foil over a large surface. For thisreason, after this first attempt, we focussed our at-tention only on the Bicron paint. After preliminary tests on the small prototypedescribed in Sec. 5.1, a larger one - 60 cm × Table 1: Attenuation length for each configuration tested in thesquare screen prototype. See text for details. covered with the same reflective paint used for thescreen and wrapped with black tape. This geom-etry enables to scan either 3 PMT modules, or 43individual PMTs per scan.In accordance with the results obtained with thesmall square prototype, we decided to use a spe-cific painting pattern. In fact, the short attenuationlength characterizing the configuration PAINT + PAINT + PAINT, points out that a layer of paint ontop of the screen is not su ffi cient to propagate thelight up to the extreme edge of a longer screen. Theadopted strategy consisted then in brush-paintingthe screen with a decreasing number of superim-posed layers of paint as a function of the distancefrom the injection edge. In this way, the coatingis thicker close to entrance, forcing the light to becarried further into the screen. In order to achievean emission as homogeneous as possible, three dif-ferent patterns have been tested:(i) 3 layers between 0 and 20 cm, 2 layers be-tween 20 and 40 cm, and 1 layer between 40and 60 cm (configuration 3-2-1);(ii) 4 layers between 0 and 20 cm, 2 layers be-tween 20 and 40 cm, and 1 layer between 40and 60 cm (configuration 4-2-1);(iii) 5 layers between 0 and 5 cm, 4 layers between5 and 20 cm, 2 layers between 20 and 40 cm,and 1 layer between 40 and 60 cm (configura-tion 5-4-2-1),where the origin is set at the light injection edge.Figure 5 shows the central longitudinal light-intensity profile, and the contour maps obtained forconfigurations (iii). The area covered within a fac-tor 1 /
2, 1 /
3, 1 /
4, 1 / Although the results obtained with the rectangu-lar screen were satisfying, exchanges with the LST6 (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/
Longitudinal distance [cm] -10-50510 La t e r a l d i s t. [ c m ] I n t en s i t y [ A DC ] Figure 5: Longitudinal light-intensity profile (top) and light in-tensity contour-map (bottom) for configuration (iii) of the rect-angular screen (color online). Contours enclose the regionswhere the light intensity is within a given percentage of themaximum intensity, I max , measured on the screen: I > < I <
50% (green), 25% < I <
33% (light blue),20% < I <
25% (blue), I <
20% (white). Light is injected intothe left edge of the screen (longitudinal distance = × camera team pushed us to decrease the ratio be-tween the length and the width of the screen, andto enlarge its surface to probe the tails of the opti-cal PSF. Both the area and the shape of the previ-ous prototype were not su ffi cient to guarantee theprobing of the PSF tails, so a 38.4 cm-diameter cir-cular screen, was built. This screen can calibrate 4modules, or 41 individual PMTs at the same time.In this configuration, inspired from that proposedfor one of the cameras of the SSTs [24, 25], thelight injection occurs all along the edge. The fish-tail light guide was replaced by two Saint-Gobainscintillating optical fibers of diameter φ / / -20 -16 -12 -8 -4 0 4 8 12 16 20 Longitudinal distance [cm]
20 16 12 8 4 0 -4 -8 -12-16-20 La t e r a l d i s t an c e [ c m ] Figure 6: Light intensity map (color online) of the 1 layer-painted circular screen. Measurements (black dots) were takenon a grid of 4 cm × light box could be a bit larger than the fibers diam-eter; (iii) the black tape at the contact point betweenthe fibers and the screen does not guarantee a per-fect adherence, leaving a small part of the fibersuncovered. In order to stem the di ff use light com-ing from (i), the region between the light box andthe screen - i.e. , where the optical fibers run - waswrapped with a foil of tedlar, a completely opaquematerial. The possible light dispersion due to (ii)was mitigated by adding a 3 cm-layer of black sili-con at the exit of the fibers from the light box. Theissue (iii) was harder to treat, and at the end a sat-isfying and reliable solution has not been achieved.Despite these technical measures, the light excesswas partially reduced, but not definitively cut out.The light intensity map obtained for the entire 1layer-painted screen is shown in Fig. 6. The lightintensity becomes increasingly lower as we moveaway both from the screen edges and the fiber en-trance. This results in a big dark region located inthe central-left zone of the screen.While the observed light distribution inhomogene-ity might be reduced with a more complicated paintpattern, the low control on the light leakage to-gether with an increased complexity of the sys-tem due to extra hardware and materials ( i.e. , op-tical fibers, tedlar, and black silicon) increases thebreakdown risk during operating condition, andcomplicates the reproducibility of the system. Forthese reasons, we decided to explore what becameour final design. The aim of the following step in the design pro-cess was the development of a new screen embed-7 (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ (a) (b) (c)(d)
Light intensity (as fraction of maximum)
I > 50%33%< I <50%25%< I <33%20%< I <25%I < 20% (e)
Figure 7: Light-intensity contour maps for di ff erent painting patterns of the octagonal screen (color online). The color code is thesame as in Fig. 5. Magenta dotted lines refer to the painting pattern (see also text for details). Black arrows indicate the light-injectionedge. Measurements were taken in correspondence with the black dots. ding both the dimension of the circular screen andthe reliability and the performance of the rectan-gular one. An octagonal screen whose length andwidth measure 40 and 42 cm, respectively, for atotal area of 1338 cm , has been built. The areaof this screen is large enough to contain 7 entirePMT modules, or 51 individual PMTs. The light-injection system is the same as for the rectangularone ( i.e. , the fishtail light guide). The screen sidefacing the mirrors has been painted with 3 layersof reflective paint. The edges have been paintedwith the same reflective paint, with the exception ofthe three farthest edges - i.e., opposite to the lightentrance - to avoid backward light-reflections. Fi-nally, all the edges are covered with black tape.Given the di ff erent geometry with respect tothe rectangular system, the coating pattern (brushpainted) towards the focal plane has been inves-tigated step by step. Starting from one homo-geneous layer, at each step, we added strips ofcoating whose position and width were determinedby analyzing the light profiles along the centrallongitudinal-axis, and the lateral one if necessary.a) 1 homogeneous layer over the entire screen surface;b) 2 layers in the first 10 cm, 1 layer on the restof the screen;c) 3 layers in the first 15 cm, 1 layer on the restof the screen;d) 3 layers in the first 15 cm, 2 layers between 15and 25 cm, 1 layer on the rest of the screen;e) 3 layers in the first 15 cm, 2 layers between 15and 25 cm, 1 layer on the rest of the screen,plus an additional layer between 0 and 30 cmalong the longitudinal distance, and between-5 and 5 cm along the lateral distance.Figure 7 shows the increasing area covered within agiven fraction of the maximum intensity measuredon the screen, for all the explored painting patterns.For configuration e), a finer grid was used to mapthe light intensity.Once satisfying results were achieved in termsof light homogeneity using the brush painting, wedecided to mount the Winston cone on the PMT, asin the real camera, to assess its impact on the mea-surements. Results are shown in Fig. 8. From acomparison with the map in Fig. 7e, the presenceof the Winston cone accentuates the impact of the8 (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ Light intensity (as fraction of maximum)
I > 50%33%< I <50%25%< I <33%20%< I <25%I < 20%
Figure 8: Light-intensity contour map of the octagonal screen(color online). The painting pattern is the same as for the con-figuration in Fig. 7e, with the addition of the Winston cone onthe PMT. Black arrows indicate the light-injection edge. Thecolor code is the same as in Fig.5. inhomogeneities on the screen. The Winston conecollects light from a defined solid angle preventingthe collection of photons coming from further re-gions of the screen at large angles. The amount ofarea covered within a factor of 1 / ffi cult to control the amount of appliedpaint. Moreover, sometimes larger drops of paintwere ejected, compromising the final homogene-ity. We abandoned this painting method because itdid not prove to be a valid alternative to the brushpainting.At this stage, another painting method was testedsince it was evident that the brush-painting tech-nique did not permit to deposit the same amountof coating in all regions of the screen. Indeed, atechnique that enables the control of the thicknessof the paint layers is needed for the pre-productionand production phases, in order to ensure a goodreproducibility of the performance of each systemthat will be mounted on several telescopes. For thisaim, a dedicated painting method (afterwards re-ferred to as dip-coating) was developed. The screenis immersed in a slim tank containing diluted paint,and then lifted up by a motorized system with aconstant speed of 2 mm / s (Fig. 9). The first trial, byusing pure paint, ended up with a too thick layer ofpaint and the presence of lumps. Hence, the paint Figure 9: Dip-coating painting method. The octagonal screenis immersed in a slim tank 2 × ×
48 cm , filled with dilutedreflective-paint. To ensure the same deposit of paint over theentire surface, the screen is lifted at a constant velocity of 2mm / s by a motorized system. has been progressively diluted with water, until anoptimal ratio (80% paint, 20% water) was achievedfor a homogeneous coverage of the screen. Thistechnique revealed a double advantage: (i) it en-ables a simplification of the final painting patternwith respect to the brush-painted solution as the ex-tra layer of configuration e) (between 0 and 30 cmalong the longitudinal axis, and between -5 and 5cm along the lateral distance) is no longer neces-sary; and (ii) it improves the light homogeneity, byextending the amount of area covered within a fac-tor of 1 /
6. Characterization of the final screen
The design process described in Sec. 5 led to thechoice of an octagonal-shape screen. It is paintedwith three layers of paint towards the mirrors and1 layer of paint on the edges, with the exceptionof the three furthest edges. The development ofthe dip-coating application method enabled an im-provement of the light homogeneity. As a matterof fact, with this method, the painting-pattern to-wards the focal plane is simplified, receding fromconfiguration e) to configuration d) (see Sec. 5.4),and the additional longitudinal layer is no longernecessary. Even with the Winston cone mountedon the PMT, the light intensity measurements im-proves significantly both in terms of area coveredwithin a given fraction of the maximum light in-tensity, and in terms of symmetry with respect tothe central longitudinal axis. Figure 11 shows the9 (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/
42 38 34 30 26 22 18 14 10 6 2 -2 -6
Longitudinal distance [cm]
24 20 16 12 8 4 0 -4 -8 -12-16-20-24 La t e r a l d i s t an c e [ c m ] Light intensity (as fraction of maximum)
I > 50%33%< I <50%25%< I <33%20%< I <25%I < 20%
Figure 10: Light-intensity contour map (color online). Light isinjected into the right edge of the screen (black arrows). Mea-surements (black dots) were taken on a grid 4 cm × =
0. Magenta dotted lines refer to the paintingpattern. The color code is the same as in Fig. 5. light intensity map measured over the screen. Thepercentages of the area covered within a given fac-tor of the maximum light intensity measured on thescreen are reported in Table 2.A study of the signal timing in terms of ar-rival time and FWHM, was performed in high-intensity regime on the final configuration (usinga PMT equipped with the Winston cone) by acquir- a b c3 layers2 layers1 layer42 38 34 30 26 22 18 14 10 6 2 -2 -6
Longitudinal distance [cm] -24-20-16-12-8 -4 0 4 8 12 16 20 24 La t e r a l d i s t an c e [ c m ] Figure 11: Light intensity map (color online). Measurementsare the same as in Fig. 10, and were acquired with the Win-ston cone mounted. Magenta dotted lines refer to the paintingpattern. Black arrows indicate the light-injection edge. Thethree red stars indicate the points where the SPE spectra wereacquired (see text and Fig. 13 for details).
Fraction Area [%] Area [cm ]1 / / / / Table 2: Percentage of the screen area covered within a givenfactor of the maximum light intensity. ing ∼ × < γ -ray event, rangingfrom 4.9 ns to 8.3 ns, with a median of 6.6 ns.The usable area per scan for the SPE acqui-sition can be evaluated according to the qual-ity of the spectrum and its statistical uncertain-ties. Three SPE spectra, containing 60,000 events,whose charge was integrated within a 20 ns win-dow, were taken at di ff erent points of the screen: ata high-, at a medium-, and at a low-intensity point,respectively, as shown in Fig. 11. A probabilitydistribution function (PDF) resulting from the con-volution of the PDF of the pedestal and n singlephotoelectron PDF, was then fit to each spectrum. The pedestal PDF is modeled by a Gaussian func-tion, while the SPE PDF is modeled by two Gaus-sians to take into account the presence of a low-charge event population [26]. The latter could bedue to electrons that, migrating from the first to thesecond dynode of the PMT, do not follow an idealtrajectory producing a loss of electrons at the firstamplification step, or to photons that produce anelectron at the first dynode instead of at the photo-catode. This kind of events lead to the formation ofa low charge component in the SPE PDF. The SPEspectra are shown in Fig. 13. The gain derived ateach position is 166 . ± . stat , 165 . ± . stat , and The time di ff erence between the first and last photo-electrons deposited in a pixel of the camera from a shower ata zenith angle of 20 ◦ shows a median of ∼ γ -ray at 100 GeV and goes up to ∼ The fit was performed with the open-source software Calin,available at https://github.com/llr-cta/calin (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ (a) (b) Figure 12: Arrival time map (a), and FWHM map (b) of the signal of the octagonal screen (color online). Light is injected into theright edge of the screen. Measurements (black dots) were taken through on a grid 4 cm × . ± . stat ADC / photoelectron, for correspond-ing I / I max light intensity of 1.0, 0.6, and 0.2, re-spectively (that corresponds to 1.57 ± stat , 1.04 ± stat , and 0.33 ± stat photoelectrons, re-spectively). The estimation of the gain does notseem to be a ff ected by the di ff erences in emissionover the screen. This means that the entire screensurface could be used for the PMT gain estimation.In-situ measurements in the camera will enable anassessment of the full capability of the SPE system.
7. Summary
The SPE system that we developed has a dualpurpose: the study of the optical PSF, both on- ando ff -axis, and the estimation of the gain of the wholephoto-detection chain. A movable target is neededfor PSF studies. The addition of a light source en-ables a robust estimation of the gain, which is ob-tained from the SPE spectrum acquired with eachPMT. The SPE calibration system exploits lightpulses injected into a PMMA screen covered by aspecial pattern of reflective paint.In order to scan the entire camera, 80 reposition-ings of the screen are needed. Operating the flasherat a frequency of O (100) Hz, and acquiring O (10 )events for each SPE, the estimated amount of timefor a full camera scan is less than three hours. The The light intensity corresponds to the mean number of pho-toelectrons of the distribution. scan can be performed during daytime provided alightproof camera enclosure.In this paper, we reviewed the design processof the screen and characterized its optical perfor-mance. In the design process, the following ele-ments have been investigated: (i) the geometry ofthe screen; (ii) the coating type; (iii) the coating ap-plication process, and (iv) coating patterns on thescreen.The final design satisfies the requirements spec-ified for the NectarCAM project, and consists ofan octagonal screen, painted with a Bicron reflec-tive paint (Saint-Gobain BC-620) using the dip-coating method. The painting pattern that opti-mizes the light homogeneity consists of 3 layersin the first 15 cm from the injection edge, 2 lay-ers between 15 and 25 cm, and 1 layer over therest of the screen. The percentage of area coveredwithin a factor of 1 / ff erent brightnesspoints, show that the entire screen surface can beused to determine the gain of the NectarCAM pho-todetection chain.The SPE system was integrated and validatedwith dedicated tests inside the NectarCAM cameraprototype mounted on the MST prototype structure11 (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ (a) (b) (c) Figure 13: SPE spectra recorded at three di ff erent light intensity points of the screen: (a) high, (b) medium, and (c) low. Thecorresponding positions on the screen are marked in Fig. 10. The spectra were acquired by using all 12 LEDs at 12.5 V, with the filterOD = in Berlin-Adlershof. Further studies will be dedi-cated to the study of the gain of the photo-detectionchain as a function of the night-sky-backgroundlevel and to the feasibility of integrating the SPEsystem in other CTA cameras. Acknowledgements
This work was conducted in the context of theNectarCAM Project of CTA. The authors are grate-ful for fruitful collaborations with colleagues fromthe CTA Consortium, in particular from the MSTStructure and NectarCAM Projects. The authorsacknowledge support from P2IO LabEx (ANR-10-LABX-0038) in the framework “Investissementsd’Avenir” (ANR-11-IDEX-0003-01) managed bythe Agence Nationale de la Recherche (ANR,France). Furthermore, we thank F. Cassol for feed-back on the expected duration of γ -ray showers inthe camera. References [1] J. A. Hinton, W. Hofmann, Teraelectronvolt Astron-omy, Annual Review of Astronomy & Astrophysics47 (2009) 523–565 (Sep. 2009). arXiv:1006.5210,doi:10.1146 / annurev-astro-082708-101816.[2] M. Actis, G. Agnetta, F. Aharonian, et al., Design con-cepts for the Cherenkov Telescope Array CTA: an ad-vanced facility for ground-based high-energy gamma-rayastronomy, Experimental Astronomy 32 (2011) 193–316(Dec. 2011). arXiv:1008.3703, doi:10.1007 / s10686-011-9247-0.[3] A. Abchiche, U. Abeysekara, ´O. Abril, et al., for the CTAConsortium, Contributions of the Cherenkov TelescopeArray (CTA) to the 6th International Symposium on High-Energy Gamma-Ray Astronomy (Gamma 2016), ArXiv e-prints (Oct. 2016). arXiv:1610.05151.[4] D. Hanna, A. McCann, M. McCutcheon, et al.,An LED-based flasher system for VERITAS, Nu-clear Instruments and Methods in Physics Research A 612 (2010) 278–287 (Jan. 2010). arXiv:0911.2015,doi:10.1016 / j.nima.2009.10.107.[5] S. D. Biller, C. W. Akerlof, J. Buckley, et al., CalibrationTechniques for Air-Cherenkov Telescopes, InternationalCosmic Ray Conference 3 (1995) 412 (Jan 1995).[6] R. Mirzoyan, On the Calibration Accuracy of Light Sen-sors in Atmospheric Cherenkov Fluorescence and Neu-trino Experiments, International Cosmic Ray Conference7 (1997) 265 (1997).[7] F. Aharonian, A. G. Akhperjanian, K.-M. Aye,et al., for the H.E.S.S. Collaboration, Calibrationof cameras of the H.E.S.S. detector, Astropar-ticle Physics 22 (2004) 109–125 (Nov. 2004).doi:10.1016 / j.astropartphys.2004.06.006.[8] G. Giavitto, S. Bonnefoy, T. Ashton, et al., for theH.E.S.S Collaboration, Performance of the upgradedH.E.S.S. cameras, International Cosmic Ray Confer-ence 35 (2017) 805 (Jan. 2017). arXiv:1708.04550,doi:10.22323 / / TNS.2002.803867.[10] M. Gaug, T. Schweizer, Martinez, et al., An AbsoluteLight Flux Calibration for the MAGIC Telescope, Interna-tional Cosmic Ray Conference 5 (2003) 2923 (Jul. 2003).[11] M. Gaug, H. Bartko, J. Cortina, et al., Calibration of theMAGIC Telescope, International Cosmic Ray Conference5 (2005) 375 (2005).[12] M. K. Daniel, M. Gaug, P. Majumdar, for the CTAConsortium, The Camera Calibration Strategy of theCherenkov Telescope Array, ArXiv e-prints (Aug. 2015).arXiv:1508.06625.[13] M. Iori, P. Majumdar, F. De Persio, et al., for the CTAConsortium, Design of a prototype device to calibratethe Large Size Telescope camera of the Cherenkov Tele-scope Array, in: 34th International Cosmic Ray Confer-ence (ICRC2015), Vol. 34 of International Cosmic RayConference, 2015, p. 954 (Jul. 2015).[14] M. C. Maccarone, D. Parsons, Gaug, et al., for the CTAConsortium, Tools and Procedures for the CTA Array Cal-ibration, International Cosmic Ray Conference 35 (2017)854 (Jan. 2017). arXiv:1709.03081.[15] J.-F. Glicenstein, M. Shayduk, for the NectarCAMcollaboration, and the CTA Consortium, NectarCAM,a camera for the medium sized telescopes of theCherenkov Telescope Array, ArXiv e-prints (Oct. 2016).arXiv:1610.04173. (cid:13) http://creativecommons.org/licenses/by-nc-nd/4.0/ [16] G. P¨uhlhofer, C. Bauer, A. Biland, et al., for the CTA Con-sortium, FlashCam: A fully digital camera for CTA tele-scopes, in: F. A. Aharonian, W. Hofmann, F. M. Rieger(Eds.), American Institute of Physics Conference Series,Vol. 1505 of American Institute of Physics Conference Se-ries, 2012, pp. 777–780 (Dec. 2012). arXiv:1211.3684,doi:10.1063 / / / / / P06008.[18] F. Arqueros, G. Ros, G. R. Elorza, et al., A techniquefor the optical characterization of imaging air-Cherenkovtelescopes, Astroparticle Physics 24 (1-2) (2005) 137–145(Sep 2005). doi:10.1016 / j.astropartphys.2005.05.009.[19] M. L. Ahnen, D. Baack, M. Balbo, et al., Bokeh mir-ror alignment for Cherenkov telescopes, AstroparticlePhysics 82 (2016) 1–9 (Sep 2016). arXiv:1605.06277,doi:10.1016 / j.astropartphys.2016.05.003.[20] A. Biland, M. Garczarczyk, H. Anderhub, et al., The Ac-tive Mirror Control of the MAGIC Telescopes, in: In-ternational Cosmic Ray Conference, Vol. 3 of Interna-tional Cosmic Ray Conference, 2008, pp. 1353–1356 (Jan2008). arXiv:0709.1574.[21] C. L. Naumann, E. Delagnes, J. Bolmont, et al.,New electronics for the Cherenkov Telescope Ar-ray (NECTAr), Nuclear Instruments and Methods inPhysics Research A 695 (2012) 44–51 (Dec. 2012).doi:10.1016 / j.nima.2011.11.008.[22] A. Bouvier, L. Gebremedhin, C. Johnson, et al., Photo-sensor characterization for the Cherenkov Telescope Ar-ray: silicon photomultiplier versus multi-anode photomul-tiplier tube, in: Hard X-Ray, Gamma-Ray, and NeutronDetector Physics XV, Vol. 8852 of Proceedings of theSPIE, 2013, p. 88520K (Sep. 2013). arXiv:1308.1390,doi:10.1117 / ff erent surface coverings, RadiationMeasurements 41 (3) (2006) 289–294 (March 2006).[24] O. Catalano, S. Giarrusso, G. La Rosa, et al., The AS-TRI SST-2M Prototype: Camera and Electronics, arXive-prints (Jul. 2013). arXiv:1307.5142.[25] D. Impiombato, O. Catalano, S. Giarrusso, et al.,Procedures for the relative calibration of the SiPMgain on ASTRI SST-2M camera, Experimental Astron-omy 43 (2017) 1–17 (Feb. 2017). arXiv:1612.02186,doi:10.1007 / s10686-016-9516-z.[26] S. Caro ff , S. Fegan, P. Jean, et al., Determination of thesingle photo-electron spectrum and gain measurement forthe Cherenkov Telescope Array camera NectarCAM, in:S. L. O’Dell, G. Pareschi (Eds.), Optics for EUV, X-Ray,and Gamma-Ray Astronomy IX, Vol. 11119, InternationalSociety for Optics and Photonics, SPIE, 2019, pp. 414 –423 (2019). doi:10.1117 /12.2530996.