PMT overshoot study for JUNO prototype detector
F. J. Luo, Y. K. Heng, Z. M. Wang, P. L. Wang, Z. H. Qin, M. H. Xu, D. H. Liao, H. Q. Zhang, X. C. Lei, S. Qian, S. L. Liu, Y. B. Chen, Y. F. Wang
aa r X i v : . [ phy s i c s . i n s - d e t ] M a y Chinese Physics C Vol. xx, No. x (201x) xxxxxx
PMT overshoot study for the JUNO prototype detector *Feng-Jiao Luo , , Yue-Kun Heng , Zhi-Min Wang
Pei-Liang Wang , Zhong-Hua Qin , Mei-Hang Xu , Dong-Hao Liao , Hai-Qiong Zhang , , Yong-Bo Huang , Xiang-Cui Lei , , Sen Qian , Shu-Lin Liu , Yuan-Bo Chen , Yi-Fang Wang , State Key Laboratory of Particle Detection and Electronics, Beijing 100049, China Institute of High Energy Physics, Chinese Academy of Sciences, Beijing 100049, China University of Chinese Academy of Science, Beijing, China Guangxi University, Nanning 530004, China
Abstract:
The quality of PMT signals is a key for large-size and high-precision neutrino experiments, while most ofthese experiments are affected by the overshoot of PMT signal from the positive HV-single cable scheme. Overshootaffects the trigger, dead time and charge measurement from a detector. For the JUNO prototype detector, we haveperformed a detailed study and calculation on PMT signal overshoot to control the ratio of overshoot to signalamplitude to ∼ Key words:
JUNO, PMT, PMT overshoot
PACS:
The JUNO[1] project is proposed to determine theneutrino mass hierarchy using a 20 kton undergroundliquid scintillator detector. As a multipurpose under-ground neutrino observatory, JUNO will also measureneutrino oscillation parameters to better than 1% accu-racy and measure neutrinos or antineutrinos from terres-trial and extra-terrestrial sources. It will provide excit-ing opportunities to address important topics in neutrinoand astro-particle physics, including supernova bursts,the diffuse supernova neutrino background, geoneutri-nos, atmospheric neutrinos and solar neutrinos. Ac-cording to its preliminary design, JUNO will distribute ∼ ′′ PMTs to reach better than 3% energy reso-lution at 1 MeV[2].Following the JUNO schedule and R&D require-ments, especially for the newly developed high quantumefficiency and large area PMTs[3], a prototype detec-tor of JUNO is planned at the Institute of High EnergyPhysics (IHEP) Chinese Academy of Science, China[4].The prototype detector is designed with 51 PMTs fromthree companies: 8 ′′ dynode PMTs and 20 ′′ dynode PMTs from Japan HAMAMATSU, 8 ′′ and 20 ′′ Micro-Channel Plate ( MCP ) PMT from Nanjing Night Vi-sion Technology Co. ( NNVT ) and 9 ′′ dynode PMTfrom Hainan Zhanchuang Photonics Co. ( HZC ). Thepositive HV scheme is used on these PMTs, and the 50-Ohm coaxial cable is carrying both the HV and signals.The scheme is widely used in many other experiments,such as Daya Bay[5], Borexino[6], Chooz[7], and DoubleChooz[8] experiments.A simplified schematic of the positive HV-single ca-ble scheme is shown in Fig.1, where a capacitor isused as decoupler to separate the signal and high volt-age. This scheme has many advantages such as lowerPMT noise, fewer cable connection and lower cost, how-ever the overshoot following a signal also causes prob-lems for charge measurements and system triggering asshown by the Double Chooz[9], KamLAND[10], SNO[11],Borexino[12], and Daya Bay experiments: it makes thetrigger system less efficient at detecting lower energyevents and distorts the charge measurement of signalsadjacent in time or after large signals from a muon cross-ing the detector. In order to overcome the disadvantages Received 31 June 2015 ∗ Supported by Strategic Priority Research Program A-JUNO,High Energy Physics Experiment and Detector R&D and NationalNatural Science Foundation of China1) E-mail: [email protected]) E-mail: [email protected]) E-mail: [email protected] c (cid:13) of overshoot, much work has been done by different ex-periments: two data dead time monitor systems wereinstalled to determine the inefficiency from overshootin Double Chooz[13]; specific triggering or data acqui-sition systems were developed in SNO, KamLAND andBorexino. At the same time, we also do not find a com-mon strategy for handling overshoot among these exper-iments. In this paper, we show a detailed study of theovershoot ratio and how to control overshoot through theoptimization within the PMT HV divider and HV-signaldecoupler. Fig. 1. Simplified schematic diagram of PMT, de-coupler, output signal and positive high voltage
A schematic of the PMT test system with positiveHV and single cable is shown in Fig.2. A LED was usedto fire PMT and PMT waveforms were sampled by os-cilloscope.
Fig. 2. The schematic view of PMT testing systemwith positive HV and one cable
The PMT and LED were put in a dark box, where theLED was driven by a program controlled pulse-generator.At the same time, a synchronized gate from the pulsegenerator was sent to a low threshold discrimination (LTD ) to trigger the oscilloscope. The high voltage wasprovided by a CAEN Mod.A1733P in CAEN SY4527crate, and the PMT anode signal after decoupler is sentto an oscilloscope with sampling rate of 1 GHz. Somesampled PMT waveforms are shown in Fig.3, where therecovery time is defined from the rising edge of PMTpulse to the overshoot amplitude recovered to 0.5% ofmaximum amplitude of signal and the ratio of overshootto signal amplitude is ∼ Fig. 3. PMT signal after decoupler: (a)1 V/divand 50 ns/div, (b) the same signal under differentscales, 200 mV/div and 100 ns/div. The signalamplitude is about 3.7 V. Overshoot is clear andthe amplitude of overshoot is about 400 mV or10.8%, it takes 700 ns to recover.
We further simplified the positive HV PMT + decou-pler + oscilloscope circuit as shown in Fig.4, where C2is the capacitor in the decoupler and R2 is the electricalload. The charge collected by the PMT anode will bereleased through the capacitors of C1 in the PMT HVdivider and C2, while the discharge of C1 and C2 willaffect overshoot. In the following sections, we will mea-sure the relationship among overshoot and C × R C × R
2, model the overshoot, and optimize the circuitto minimize overshoot to meet the requirement of theJUNO prototype.
Fig. 4. Simplified circuit of decoupler: differential circuit010201-2hinese Physics C Vol. xx, No. x (201x) xxxxxx
The signal overshoot is a result of the dischargingof capacitors with resistors. Here we measured the rela-tionship between the overshoot and R1, where R1 rangedfrom 1 Ω to 10 kΩ following the scheme in Fig.4 ( all othercomponents are shown as in the figure ). The results areshown in Fig.5 and Table 1.
Fig. 5. (a):The ratio of overshoot to signal ampli-tude as a function of R1; (b): The time of signalbaseline recovering to 0.5% of maximum ampli-tude of signal as a function of R1.
From Fig.5(a), we learn that the ratio of overshoot tosignal amplitude decreases when R1 increases. Accord-ing to Fig.5(b), the time of the signal baseline recoveringto 0.5% of the signal maximum increases when R1 in-creases where the recovery time is consistent with thetime constant of R × C
1. The recovery time shows aknee around R1=250 Ω, above which the overshoot ratiodecreases more slowly and the recovery time of the signalbaseline almost keeps nearly constant around 1 µ s whichis consistent with the time constant of C × R
2. And wetested all 5 types PMTs used in JUNO prototype, andall got consistent results.From the data ( shown in Table 1 ) and calculation,we know that when the value of R × C << R × C R × C R × C
2. When R × C >> R × C
2, the discharge of R × C R × C Table 1. The measurement results of overshootvs. resistor R1.
R1/Ω Ratio of overshoot/% Recover time/ns1 62 505 26 24020 13 42050 10 50075 8.0 600125 5.3 640150 4.6 750200 3.2 800225 2.9 900234 2.7 950250 2.69 980350 2.39 1000400 1.95 1020500 1.56 1020600 1.11 10401000 1.02 110010000 1.0 1100
Following the understanding of overshoot that it isthe discharge of the capacitors in the PMT HV di-vider and decoupler, we can simplify the system furtheras shown in Fig.6, where we can model the output ofPMT+decoupler as:
Fig. 6. Further simplified circuit V o ( t ) = V i ( t ) × h ( t ) (1)Where V i ( t ) is the PMT output which is the systeminput, h ( t ) is the response model of the PMT HV divider+ decoupler +oscilloscope and V o ( t ) is the final waveformviewed by an oscilloscope.We can use a Fourier transformation to analyze thewhole process shown in Fig.7. Adopting the transformation from time to frequencydomain, we can get the output signal of the system de-scribed as: V o ( w ) = V i ( w ) × H ( w ) (2)where H ( w ) is the response of the system in the fre-quency domain. According to the simplified circuitshown in Fig.6, it can be expressed as: H ( w ) = jw/ ( jw + 1 /τ ) (3)where τ is the time constant of the differential circuitwhich is 22 nF ×
50 Ω ∼ V i ( t ) as anexponential pulse, where τ i is the time constant of theanode output, which is usually <
12 ns. V i ( t ) = − Q/C i × exp( − t/τ i ) (4)For equivalent calculation, we assume a positive pulseinstead of negative pulse. Then we can get the input inthe frequency domain, V i ( w ) = Q/C i × / ( jw + 1 /τ i )) (5)Combining equations (1), (2), (3) and (4), we obtain, V o ( w ) = V i ( w ) × H ( w ) = Q/C i × ( jw ) / (( jw +1 /τ ) × ( jw +1 /τ i ))(6)By means of the inverse Fourier transform, the out-put of system can be obtained: V o ( t ) = Q/C i × (1 / ( τ i − τ )) × ( τ i exp( − t/τ ) − τ exp( − t/τ i ))(7)Eq.7 is the output of the circuit when the input is anexponential pulse. According to the equation, the outputwill have an overshoot, and when τ >> τ i , the overshootratio can be expressed as: V − V M = τ i τ (8)where the V − is the amplitude of the overshoot and the V M is the maximum amplitude of the signal. Thus, thevalue of V − /V M is about 1.1% for τ i ∼
12 ns and τ ∼ ∼
8% for τ ∼
140 ns.For the JUNO prototype, we selected a single ended50 Ω matching design with R1= 10 kΩ and C1= 4.7 nFand C2= 22 nF. The final output waveform is shownin Fig.8. The overshoot of the signal is about 1% and isconsistent with our calculation with the overshoot model.
Fig. 8. PMT signal after decoupler: (a)1 V/divand 50 ns/div, (b) the same signal under differ-ent scales,100 mV/div and 100 ns/div. The sig-nal amplitude is about 3.0 V. Overshoot is smalland the amplitude of overshoot is about 30 mVor 1.0%, it takes more than 1us to recover.
The PMT HV divider plays a crucial role in the PMTperformance[15]. The timing and linearity of the PMTresponse are good parameters for checking the PMT HVdivider design. We have already reduced the overshootfrom ∼
10% to ∼
1% as discussed in the previous sections.We tested all 5 types PMTs used in JUNO prototype,and all got consistent results. Here we just selected HZCXP1805 PMT as an example to show the typical results.
With the optimized HV divider for a HZC XP1805PMT, we measured the rise time and fall time with thesystem shown in Fig.2.We get the waveforms of single photoelectron ( SPE) signals using an LED producing photoelectrons at anaverage occupancy of 10%. Then we can record the wave-form by oscilloscope with 1 GHz sampling, as shown inFig.9.
From the measured waveform, we know that thePMT with our updated HV divider still has its fast timeproperties and reaches our expectation.
The linearity of the PMT can be obviously influencedby the PMT HV divider design. The definition of pulselinearity is the ratio of input and output photoelectronsin pulse operation mode. The measurement scheme isshown in Fig.10.
Fig. 10. The schematic of linearity measurement.
A pulse generator can drive two blue LEDs and flashthem at the same time. The light intensity of each LEDcan be tuned separately to cover the whole PMT dy-namic range. The PMT can see a sequence of lightpulses: a light pulse ( A ) from LED1, a light pulse (B ) from LED2, and a sum of light pulses ( C ) from thetwo LEDs flashing simultaneously. At the same time, aFADC system with 1GHz sampling will record the wave-forms corresponding to the three different light pulses.If the PMT is ideally linear, we can have, C = A + B (9)In reality as known, the PMT response has non-linearity effect. Then the deviation of the linearity isdefined as, N onlinearity = C − ( A corrected + B corrected ) A corrected + B corrected (10) Where A corrected ( B corrected ) is the corrected A ( B )according to measured nonlinearity effect in lower inten-sity range. The measurement results are demonstratedin Fig.11. The figure shows that the 5% deviation canreach around 700pe. The measured linearity of the PMTis satisfied our requirements and reached the same levelwith previous measured R5912[16]. Fig. 11. Nonlinearity as function of the number ofthe photoelectrons of the anode output pulse.
In this work, we confirmed a large overshoot of thePMT output at the level of 10% as observed by the DayaBay experiment[14], and made clear that the PMT over-shoot originates from the discharge of the decouplingcapacitor of the HV-signal decoupler and the capaci-tor in the HV divider. The design of HV divider anddecoupler has been optimized to reduce the overshootfor charge measurements. The value of R × C R × C The project was supported by Strategic Priority Re-search Program A-JUNO, High Energy Physics Experi-ment and Detector R & D and National Natural ScienceFoundation of China. References :030401 (2016)3 X. C. Lei, Y. K. Heng, S. Qian et al, Chin. Phys. C, (2) :26002-026002 (2016)4 JUNO collaboration, http://arxiv.org/abs/1508.071665 F.P. An et al, Nucl. Instrum. Meth. A, : 78—97 (2012)6 G. Alimonti et al, Nucl. Instrum. Meth. A, (3) : 568—593(2009)7 A. Baldini, C. Bemporad, E. Caffau et al, Nucl. Instrum. Meth.A, (1-2) : 207—221 (1996)8 C. Bauer, E. Borger, R. Hofacker et al, Journal of Instrumen- tation, : P06008 (2011)9 Y. Abe et al, Phys. Rev. D, (5) : 052008 (2012)10 S. Abe et al, Phys. Rev. Lett, : T01003 (2013)14 S. Jetter, D. DWYER, W. Q. Jiang, Chin. Phys. C, (8) :733-741 (2012)15 Hamamatsu Photonics K. K, Photomultiplier Tubes : Basicsand Applications, Third edition (2007)16 W. L. Zhong, J. C. Liu, G. C. Yang et al, Chin. Phys. C, (05) : 481-486 (2007)(05) : 481-486 (2007)