Trigger Merging Module for the J-PARC E16 Experiment
12 Jun 2018 Kyoto Univ.A, RIKENB, RCNPC, KEKD, Tohoku Univ.E
M. IchikawaA, B, M. NarukiA, S. YokkaichiB, T. N. TakahashiC, K. AokiD, S. AshikagaA, E. HamadaD, R. HondaE, Y. IgarashiD, M. IkenoD, K. OzawaD, H. SendaiD, K. N. SuzukiA, M. TanakaD and T. UchidaD
�1
Outline �2◯J-PARC E16 Experiment ・Detectors ・Trigger System
◯Trigger Merging Module(TRG-MRG) ・Hardware ・Firmware
◯Performance Test ・Time Resolution ・Latency ・Transfer Efficiency
◯Summary
J-PARC
J-PARC
30 GeV Proton Synchrotron Circumference: ~1.6 km
※The E16 experiment will start 2019
Japan
J-PARC E16 Experiment
�3
Detectors �4
and further, the first data of momentum dependence of the spectral change, with the 6 timeshigher statistics than that of E325 collected, as shown in this section. In addition, we alsomeasure the yield and kinematical distribution of vector mesons, and background conditionsin the limited detectors in this stage. Based on the information, we could blush up thedetailed plan of the next stage.
With additional budget of about 5 Oku yen, we will able to prepare the full acceptance,which consists of 26 modules. (Cost for new module is 0.26 Oku yen/module including readoutelectronics.) The configuration is illustrated in Fig. 8. We would like to run 320 shifts (RUN2) with the configuration, and take physics data. We will be able to do systematic study ofthe in-medium spectral change of vector mesons.
Here, we would like to request approval for RUN 0 and RUN 1.The expected yield of vector mesons expected for the different configuration and different
lengths of the data taking is listed in Table I. For the estimation, we assume the usage of a400 µm -thick Carbon target and two 80 µm -thick Copper targets simultaneously, locatedin-line on the beamline. Thus the total interaction length is 0.2 % and the interaction rate is10 MHz at the target with the beam of 1×1010 protons per 2-sec pulse. Other numbers usedfor the yield estimation is summarized in Table II.
FIG. 7: Detector configuration with 8 modules.FIG. 8: Detector configuration with 26 modules.(RUN 2)
RUN beam time configuration target φ ωRUN 0 9 shifts 6 + 6 + 2 + 2 Cu 460 2400RUN 0’ 9 shifts 8 + 8 + 8 + 8 Cu 840 4400RUN 1 160 shifts 6 + 6 + 2 + 2 Cu 8200 42000RUN 1’ 160 shifts 8 + 8 + 8 + 8 Cu 15000 (1700)RUN 1’ 160 shifts 8 + 8 + 8 + 8 C 12000 (1500)RUN 2 320 shifts 26 + 26 + 26 + 26 Cu 69000 (12000)
KEK-PS E325 Cu 2400 (460) 3200
TABLE I: Numbers of φ’s and ω’s expected for different configurations and lengths of the beam time,compared to the numbers obtained by the KEK-PS E325 experiments. Numbers in parentheses arefor mesons with βγ < 1.25. Four numbers interleaved with “+” in the configuration column describesthe numbers of the modules for the four spectrometer components : SSD, GTR, HBD, and LG,respectively.
10
Proton Beam
Hadron Blind Detector(HBD)
Lead Glass Calorimeter(LG)
GEM Tracker (GTR)×3
Target ChamberSilicon Strip Detector (SSD)
Flight Path Detection -> Momentum Measurement
Electron Identification
Using 4 types detectors
◯Readout ・112,996 ch ・Taking waveform data ・Waveform sampling time: 1 ns ・Waveform buffering time: 2 μs
◯Trigger ・Using GTR3, HBD, LG ・2,620 ch ・
Trigger System �5
GTR3
HBD
ASD
ASD
LG DRS4 ADC/TDC
Trigger Merging Module
GTR3 ASDTrigger Merging Module
GTR3 ASDTrigger Merging Module
GTR3 ASDTrigger Merging Module
Trigger Merging Module
HBD ASDTrigger Merging Module
HBD ASDTrigger Merging Module
HBD ASDTrigger Merging Module
Trigger Merging Module
LGDRS4
ADC/TDC
Trigger Merging Module
LG DRS4 ADC/TDC
Trigger Merging Module
LG DRS4 ADC/TDC
Trigger Merging Module
Trigger Decision Module
Belle-2 UT-3
Trigger Distribution Module
Belle-2 FTSW
to Readout Modules
TRG-MRG ~15 modules
Detector, Discriminator ・2,620 ch ・
Trigger Merging Module (TRG-MRG)
�6
TRG-MRG: Hardware �7
SFP+: optical links ×8
FPGA: Xilinx Spartan3 50AN-4
RJ45: for Remote JTAG
RJ45: FTSW
Connector to Mezzanine ×2
NIM I/O ×212 V Power Supply
DC/DC
Main Board Mezzanine Card128 ch LVDS Receiver
LVDS->1.8V LVCMOS Converter
125 MHz Crystal Oscillator ×2
FPGA: Xilinx Kintex7 160T-2
250 MHz
TRG-MRG: Firmware �8
500 MHz DDR Deserializer
Edge Detecter
Delay Control(up to 1,024 ns)
Hit Buffer
FIFO
Aurora8b10b ×4 lanes
Discriminator 256 ch
Hit Buffer
Hit Buffer
FIFO FIFO FIFO
SFP+ SFP+ SFP+ SFP+
Using Vivado IP Core(ISERDESE2) 1 GHz × 1 bit -> 250 MHz × 4 bit
Buffering maximal 8 hits for 64 ns in each 64 ch
32 bit × 5 cycle / 64 ns -> 2.5 Gbps
6.25 Gbps optical transceiver
Link-layer protocol for high-speed serial communication
1 ns Sampling, 256 ch Multi Hit TDC + 6.25 Gbps 4 GTX Transceivers
TDC
Transceiver
For clock domain crossing
LVDS to LVCMOS Converter Mezzanine
Hit Buffer
156.25 MHz
※Implemented by Vivado2017.2
Performance Test
�9
Time Resolution �10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
tin[ns]
ΔT[ns]
5.10
h0Entries 1256244Mean 1189Std Dev 0.1338
1188 1188.5 1189 1189.5 1190 1190.5 1191 1191.5 11920
200
400
600
800
1000
1200
310× h0Entries 1256244Mean 1189Std Dev 0.1338
TIMELAG
1189 11901188 1191時間差[ns]
カウント数
5.11 3bin :
y =√x(1− x)/2 2
62
Function Generator Delay TRG-MRG
1 kHz Pulse width: ~50 ns
Discriminator
5.1.4
5.2 5.3
DRS4 500MHz
(DSOX4054A[30]) DRS4
h0Entries 1463002Mean 16.05Std Dev 0.2212
15 15.5 16 16.5 17 17.5 18 18.5 19Time Lag[ns]
0
200
400
600
800
1000
1200
1400310×
Cou
nt h0Entries 1463002Mean 16.05Std Dev 0.2212
5.2 2ch
54
Distribution of time difference between 2 channel
Distribution of time resolution
※tin: the remainder of (time difference) / LSB
Red line: Quantization error
Time resolution < 0.35 ns in all tin region Good agreement with expected error
Time Difference[ns]
Latency �11
Edge Detect
Delay Control
Hit Buffer
FIFO
Aurora8b10b ×4 lanes
Hit Buffer
Hit Buffer
Hit Buffer
FIFO FIFO FIFO
SFP+ SFP+ SFP+ SFP+
TDC
Optical CableSFP+ SFP+ SFP+ SFP+
500MHz DDR Deserializer
Transceiver
Aurora8b10b ×4 lanes
◯Latency of TDC section From deserializing LVDS signals to transmitting data to transceivers with no delay
◯Latency of transceiver section From inputting data to FIFO to outputting data from Aurora through optical cable
Encode Serialize
Deserialize Decode
Latency: TDC Section �12
※Using circuit simulator in Vivado2017.2
レイテンシ 15
回路シミュレーターを用いて確認
Receive discriminator output
Transmit hit data
64 ns 179 ns
64 ns0 ns
Detect leading edges -> Buffering for 64 ns
Latency in TDC section is
Entries 3.5e+07Mean 288Std Dev 3.692
285 290 295 300 305 310 315 320
410
510
610
710
Entries 3.5e+07Mean 288Std Dev 3.692
latency {event
h0Entries 132808Mean 2.094RMS 1.491
0 2 4 6 8 10 120
5000
10000
15000
20000
25000
30000
35000h0
Entries 132808Mean 2.094RMS 1.491
nhit
h1Entries 132808Mean 3.615RMS 1.949
h1Entries 132808Mean 3.615RMS 1.949
Transfer Efficiency of the Merging Criteria �14Merging Criteria: Allow maximal 8 hits for 64 ns in 64 ch
Beam Rate[GHz]6 8 10 12 14
TRG
-MR
G R
elay
ing
Effic
ienc
y0.94
0.95
0.96
0.97
0.98
0.99
1
5 GHz 10 GHz
99.95 %
98.75 %
Discarded hit
Multiplicity for 64 ns in 64 ch
Beam intensity dependence of the transfer efficiency
Beam rate
99.95% transfer efficiency in expected situation Enough to use in the experiment
Blue: Beam Rate = 5 GHz Red: Beam Rate = 10 GHz
◯Simulate expected situation in the experiment with Geant4 ・Beam Rate: 1×1010 /pulse (5 GHz) ・Single rate:
Summary �15・Develop trigger merging module(TRG-MRG) for the J-PARC E16 Experiment
・Detect leading edges from discriminator output by FPGA and transmit serialized data to trigger decision module by SFP+
・1 main board + 2 mezzanine cards - Mezzanine: Receive 128 ch/card LVDS signal + Convert LVDS to LVCMOS format - Main: 1 ns sampling, 256 ch Multi Hit TDC + 6.25 Gbps optical transceiver ×4
・Required performance is achieved - Time Resolution:
pφ
e+
e-NucleusBranching Ratio ~3×10-4
Di-lepton Measurement
Exp. Method
J-PARC E16 Experiment �17
/ ndf 2χ 87.25 / 51
Prob 0.001181
0.85 0.9 0.95 1 1.05 1.1 1.15 1.20
200
400
600
800
1000
/ ndf 2χ 87.25 / 51
Prob 0.001181
/ ndf 2χ 37.12 / 38
Prob 0.51
0.85 0.9 0.95 1 1.05 1.1 1.15 1.20
200
400
600
800
1000
/ ndf 2χ 37.12 / 38
Prob 0.51
/ ndf 2χ 88.39 / 51
Prob 0.0009053
0.85 0.9 0.95 1 1.05 1.1 1.15 1.20100200300400500
600700800900
/ ndf 2χ 88.39 / 51
Prob 0.0009053
/ ndf 2χ 37.83 / 38
Prob 0.4773
0.85 0.9 0.95 1 1.05 1.1 1.15 1.20100200300400500
600700800900
/ ndf 2χ 37.83 / 38
Prob 0.4773
/ ndf 2χ 87.41 / 51
Prob 0.001139
0.85 0.9 0.95 1 1.05 1.1 1.15 1.20
100
200
300
400
500
600 / ndf 2χ 87.41 / 51
Prob 0.001139
/ ndf 2χ 30.55 / 38
Prob 0.7997
0.85 0.9 0.95 1 1.05 1.1 1.15 1.20
100
200
300
400
500
/ ndf 2χ 30.55 / 38
Prob 0.7997
/ ndf 2χ 88.95 / 51
Prob 0.000795
0.85 0.9 0.95 1 1.05 1.1 1.15 1.20
20
40
60
80
100
120
140
160 / ndf 2χ 88.95 / 51
Prob 0.000795
/ ndf 2χ 39.18 / 38
Prob 0.4167
0.85 0.9 0.95 1 1.05 1.1 1.15 1.20
20
40
60
80
100
120
140
160 / ndf 2χ 39.18 / 38
Prob 0.4167
Simulation results : Run1, Cu target
•さらに、βγで4つのregionに分けてFitをした結果 •全てのregionで有意なexcessが見えていて、βγ-dependenceが確認できる
10
�� < 1.25 1.25 < �� < 1.75 1.75 < �� < 2.5 2.5 < ��
Fit for the
whole region
Fit except for
the excess region
Mass [GeV/c2]Mass [GeV/c2]Mass [GeV/c2]Mass [GeV/c2]
Counts /[6.67 MeV/c2 ]
Counts /[6.67 MeV/c2 ]
[GeV/c2]
Expected Spectrum
W. Nakai, JPS2017
Mass modification due to nuclear effect
In vacuum
J-PARC
30 GeV Proton Synchrotron Circumference: ~1.6 km
Motivation: Restoration of chiral symmetry breaking under nuclear density Method: Invariant mass spectroscopy of φ( ) in nuclear medium Facility: Japan Proton Accelerator Research Complex(J-PARC) High momentum(High-p) beam-line (Completed by FY2019) Beam: 30 GeV proton, ~1×10^10 /pulse(~2 s)
s̄s
High-p beam-line
Trigger Concept �18
Beam
e-
e+ GTR3HBD
LG
DetectDetect
>50°
◯Target e+ e- pair from φ
◯Main Back Ground ・Miss ID of charged π ・e+e- from π0 - π0 -> 2γ, γ Conversion - π0 -> γe+e- (Dalitz Decay)
◯Method ①Detect e+/e- track by 3 detector’s coincidence (Single rate: To avoid charged π ②Require the opening angle > 50° -> To reject e+e- from π0
◯Expected Trigger Condition ・Trigger rate: ~1 kHz ・φ -> ee Surviving Ratio: ~75 %
Top View of Detectors
πDetect
Not Detect
Integral Non Linearity(INL) �19
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200
Input Time Lag[ns]
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
Out
put T
ime
Lag[
ns]
/ ndf 2χ 29.17 / 35Prob 0.745p0 0.003807± 0.7259 p1 06− 5.194e± 1
/ ndf 2χ 29.17 / 35Prob 0.745p0 0.003807± 0.7259 p1 06− 5.194e± 1
5.3
5.2
0.02 ns ( : 5.2)
1± (5× 10−6)
5.2
GTR3 ASD
500 ns 500 ns
2 µs
55
5.2.1
1LSB 1ns
INL 1LSB
5.2.2
2
INL
5.2.3
5.4 5.3
10 210 310Input Time Lag [ns]
0.06−
0.04−
0.02−
0
0.02
0.04
0.06
Res
idua
l [LS
B]
Integral Non Linearity
5.4
5.3
0
[−0.035LSB, 0.035LSB] 1LSBINL 1ns
56
Residual[LSB]
Input Time Lag[ns]10^1 10^2 10^3
Relation between Input/Output Time Lag Distribution of Residual
Red Line: At+B Fit
Function Generator Delay TRG-MRG
1 kHz Pulse width: ~50 ns
Discriminator
INL = [-0.04, +0.04] LSB with time lag ~0-2000 ns No bad influence
Differential Non Linearity(DNL) �20
DNL = [-0.022, +0.022] LSB No bad influence
Deserialize 1 GHz -> 250 MHz -> Clock interval has 4 ns periodicity -> Check each clock width from distribution of input clock signal
80.008 ns Clock TRG-MRG
Time%8[ns] Time%4[ns]
Distribution of clock edge timing
4 ns Periodicity
Minimum Pulse Width �21
Detect narrow signal (1.0-1.3ns) with 100% efficiency Enough to use in the experiment
Double Pulse Separation
1.0-1.3 ns
Input Signal
Discriminate adjacent 2 signals (2.5-2.8ns) with 100% efficiency Enough to use in the experiment
2.5-2.8 ns
Input Signal
※Those results are limited by test setup
Detect
Discriminate
Threshold
Threshold