High Energy Beam Test and Simulation of
Silicon-Tungsten Calorimeter Test Module
Shinwoo Nam (Ewha Womans University)
On behalf of
Ewha Womans University: S.J. Baek, H.J. Hyun, S. Nam, I.H. Park, J. Yang
Korea University: J.S. Kang, S.K. Park, J.H. Choi
Kyungpook National University: Y.D. Oh, K.H. Han, D.H. Kim, J.S. Seo, U.C. Yang
Sungkyunkwan University: I.T. Yu, Y.P. Yu
Yonsei University: B.S. Jang, S.H. Jeong, J.H. Kang, Y.J. Kwon
Introduction
• Previous studies have shown that the precision level of jet measurement
required for ILC physics can be achieved by reconstructing all single
particle showers in a jet.
• This leads us to two requirements for calorimeter : small Moliere radius
and fine granularity.
• Silicon-Tungsten Calorimeter is a natural candidate of EM calorimeter
meeting the conditions.
Content
• Development of Silicon Sensor
• Electronics and Mechanics of Test Module
• MC Simulation
• Beam Test Result
• Summary
Silicon Sensor (Pixellated PIN Diode)
Guard Ring
Pixels(Signal)
SiO2
p+
380㎛
N-type silicon wafer of 5 ㏀
20um
Al
60um
•Fabricated on 380um 5’ wafer
•A Sensor Size : 6.52*5.82 cm2
(including 3 guard rings )
•Pixel array : 4*4 matrix
1.55 * 1.37 cm2 each
• DC coupled
•Full depletion voltage : 90V
•Leakage current level : about 3 nA
per pixel at full depletion voltage
3 Guard Rings
Process of Silicon Fab, Sawing, Bonding
Clean wafer
Oxidation
Wafer -> Fabricated PIN diode matrix
Mass Production Fabrication made at
SENS Technology (www.senstechnology.co.kr)
N+Diffusion
Cover with
photoresist
Expose
through
mask
Develop
Etch, Stip
Sawing / attach Kapton tape
Kapton tape has patterned Cu wiring(50um) on
it for readout
Wire bonding
For wire boning to the diode pixel, Al wire with
diameter 25 um was used. Recently we added
one more wire to reduce risk of bonding failure
P+
Implantation
Anneal
Metallization
Fabrication process
Glob Top (DCE, DP100)
For protection of bonded wire.
It is important to put the glob top in
Vacuum to remove the air bubbles in glue.
Capacitance Measurement
CV
1/c2
1.20E+18
ED3_10_all
Capacitance(F)
1.00E+18
ED3_11_all
ED3_12_all
8.00E+17
ED3_13_all
6.00E+17
ED3_14_all
ED3_17_all
4.00E+17
ED3_19_all
ED3_2_all
2.00E+17
ED3_20_all
160
150
140
130
120
110
100
90
80
70
60
50
40
30
20
10
0.00E+00
Reverse bias voltage (V)
Full depletion voltage for 5kOhm wafer sensor: about 85-90V
Applied 100V because of variation in the thickness and
resistivity of wafers
Leakage Current Measurement
IV
ED2_4_1
ED2_4_2
1.00E-07
ED2_4_3
ED2_4_5
(10nA)
ED2_4_6
1.00E-08
ED2_4_7
ED2_4_8
ED2_4_9
1.00E-09
ED2_4_10
ED2_4_11
ED2_4_12
Reverse bias voltage(V)
160
150
140
130
120
110
100
90
80
70
60
50
40
30
20
1.00E-10
10
Leakage current(A)
ED2_4_4
ED2_4_13
ED2_4_14
ED2_4_15
ED2_4_16
~3nA per pixcel at full depletion voltage !
Close to 90% yield with quality cut of 20nA/pixel at 100V !
S/N Ratio Measurement with Sr-90 source
(use of single channel very low noise preamp)
Dark box
Photodiode
Pb
sensor
Pb
Beta (90Sr) source
Trigger
Photodiode
Discriminator
Gate Generator
Shaping AMP
S/N ~ 120
PreAmp
PreAmp for sensor
Frontend Readout with CR1.4 chip
16 pixels
Adjustable
MOS Resister
3000
Cc
SA
Equation: Y = -5.66 + 0.98*X
2500
Cfs
Cf
CSA
Calibration
MUX
Self
Trigger
T/H
Ct/h
V to C
MUX
ADC
Gain
Output
Buffer
T/H Output (mV)
Adjustable
Reset
2000
1500
1000
500
0
0
500
1000
1500
2000
2500
3000
Input Charge (fC)
•Developed for the Pamela Experiment
•16 channels of charge inputs (integrating the
charge pulses -> DC levels)
•Gain: 1mV/fC
•Dynamic Range: to 4000 MIPs
•up to 150 pF capacitance with leakage
currents as high as 100 nA. It measures
charge from 2.2 fC to 9 pC.
•Noise ~ 5000 e
•Power: 0.3 mW/ch
•The outputs of the T/H circuits are multiplexed
to a common output buffer that is capable of
driving a load of 1k and 100 pF.
•The output of the chip swings from -3V to 4V
Gain Linearity Test
Using charge calibration
Function of chip
CR1.4 chip handles a 16-ch Si sensor
Frontend board with 7 CR chips
Digital Electronics : ADC, Control, Power Board
ADC: MAX 1133
Power Control
FPGA
DAQ
board
PC
DC Voltage
High Voltage
ACP Board
ADCs
• Sampling Speed : 200ksps
(200ksps X 16bit = 0.4Mbyte/s)
• Resolution : 16bit (65536 Levels)
Frontend Board
Integration Test of Electronics and DAQ
readout speed : 0.1 msec for full readout
ADC : 16 bits
Data IO, Command,
Calibration Boards
ADC, Control Board
Total 640 readout channels
Tungsten and Mechanics
Tungsten
thickness : 3.5 mm (= 1 X0)
Size 65.5 mm X 57.5 mm ( ~ sensor size)
Test Module : 20 layers stacked
Frontend board
Mount holes
Aluminum Support of a Layer
Thickness of an Assembled Layer
Aluminum
1.5 mm
Sensor and Readout
10 mm
Tungsten
3.5 mm
Connector
Pcb
Diode
2.7 mm
1.7 mm
1.15 mm
Shielding board
Capacitor
Resistor
1.4 mm
0.65 mm
CR 1.4 chip
2.45 mm
15 mm
1mm
1mm inactive gap between sensors
Silicon
Sensor
131mm X 115mm
Frontend
Board
32 pixels
in a layer
Layers of Si sensors
and Tungstens
Frontend readout boards
Beam
Direction
Digital
and Control
Boards
Summary of Our Test Module
•
•
•
•
Geometry
Total 20 layers = 20X
with uniform layer thickness
Shower sampling at 19 layers
with 2 sensors each layer.
1mm gap between sensors
Aligned beam center to the center of a sensor
1mm inactive gap
Effective RM :~ 45mm
from volume ration of material
RM
131mm X 115mm
-> insufficient transverse shower containment
No action taken for cooling the detector.
Temperature level during test : ~35 to 40 deg
Geant4 Simulation
Geant4.7 with full official high energy physics list,
Detector geometry same as real test module,
Fixed beam gun without spread,
Detector responses in terms of pure energy loss
(no noise and digitization)
<30GeV, e-, for 1/500 event>
E(MeV)
Electron Longitudinal Shower Profiles
100GeV
50GeV
Radius(mm)
10GeV
RM is about 45mm
Geant Simulation
total energy loss in silicon layers
vs. electron beam energy
MeV
dE/E vs. E
Fitted curve to…
18 % / √E
GeV
GeV
<20GeV>
<50GeV>
CERN Beam Test
Steps of Beam Test
1. Tune trigger time delay
beam
2. Align detector by using
movable table
under the our detector
3. MIP calibration of all channels
(using hadron beam (less
spread) after removing all
tungstens)
Thanks A. Malinine for the test beam line control
Beam Test : CERN SPS H2 beam line
for a week till Sep. 7 2004
Beam cycle 18.0 sec with 4.8 sec spill time
beam line focus & existing trigger scintillators
give beam spread of ~1 cm diameter
Beam focus worse in muon beam
4. Data Run
(electron
150,100,80,50,30,20,10 GeV
hadron 150 GeV
muon 150 GeV)
random trigger mixed in the
runs for pedestal monitor
Channel Scan for MIP calibration
Scanned over all 640 channels
with 100 GeV hadron Beam
(no tungsten)
Pedestal :
Gaussian Fit
Mean : 5206.9
Sigma : 7.2
an example of a sensor with all good pixels
Signal :
Landau Fit
Peak : 5243
S/N = 5.2
ADC Counts
Detector Response to Different Particles
Random Trigger events (total pedestal)
50 GeV Electron
50 GeV pion
Online Shower Profile Monitor
Pedestal subtracted
First Analysis :
sum ADC counts of
all channels
150 GeV Muon
No rejection of dead, noisy
channels, No gain
calibration applied
Total ADC of an event / 640
Detector Response to Different e- Energy Shower
150 GeV
100 GeV
80 GeV
Readout
Pedestals
from
Random
Trigger
50 GeV
30 GeV
20 GeV
10 GeV
Total ADC of an event / 640
Calorimeter Calibration
Total ADC above pedestal / 640
Preliminary
Straight Fit Line
1GeV <--> 4.2 * 640 ADC Counts
Linear response, No saturation
Electron Energy in GeV
Energy Resolution
Preliminary
dE / E (%)
 Geant4 simulation of this
setup taking into account
only shower leakage gives
18%/√E.
Fit curve : 29%/√E
 Simulation with dead
channels (~2%), noisy
channels (~10%), ADC
unstable(~10%) : 21%/√E.
 Further analysis needed
to study the influence of
beam spread, insufficient
gain calibration, and
readout channel crosstalks.
Electron Energy in GeV
Summary
• High-quality Silicon pixel sensors were successfully developed and produced.
- the yield close to 90% (better than initial expectation)
– excellent quality, typically Id <=10nA/cm2
• Si-W Test Module was built and exposed to the high-energy beams
- Obtained typical performance as an EM calorimeter in the detector
responses to the different beam energy and particle type.
- Preliminary result of electron energy resolution : 28%/√E,
- MC result : 18%/√E for perfect readout
21%/√E taking into account of noisy and bad channels
- Under study for the effect of gain calibration, beam spread,
and cross-talks in readout electronics
• Further tests of silicon sensors with more practical integration condition is in
preparation.