Dr Renato Turchetta
Instrumentation Department
CMOS Monolithic Active Pixel Sensors(MAPS) for the ILC
2
RAL, 16/2/2005
OutlineIntroduction on MAPS
MAPS for Particle Physics
Parametric test sensors RAL_HEPAPS
Source results
Irradiation results
Flexible APS (FAPS) for ILC. Concept and
source results
MAPS for ILC
3
RAL, 16/2/2005
CMOS (Monolithic) Active Pixel Sensor (MAPS)
Standard CMOS technology
all-in-one detector-connection-
readout = Monolithic
small size / greater integration
low power consumption
radiation resistance
system-level cost
Increased functionality
increased speed (column- or
pixel- parallel processing)
random access (Region-of-Interest
ROI readout)
Column-parallel ADCs
Data processing / Output stage
Rea
dou
t co
ntr
olI2
C
con
trol
(Re)-invented at the beginning of ’90s: JPL, IMEC, …
4
RAL, 16/2/2005
CMOS sensors: camera architecture
Column-parallel ADCs
Data processing / Output stage
Column-parallel ADCs
Data processing / Output stage Cam
era
con
tro
l
Rolling shutter Snapshot
Integration (exposure) and readout are interleaved
Integration time given by time between two readouts
Simultaneous integration for all pixels, followed by the readout
Cam
era
con
tro
l
5
RAL, 16/2/2005
Metal layers
Polysilicon
P-Well N-Well P-Well
N+ N+ P+ N+
CMOS for
100%
efficient
detection
of charged
particles
Dielectric for insulation and passivation
Radiation
--
--
--
- ++
+++
++
- +- +- +
P-substrate
P-epitaxial layer
Potential barriers
epi
sub
N
Nln
q
kTV
Concept first proposed in 1999, and published in NIM in 2001 (R. Turchetta et al.)
6
RAL, 16/2/2005
MAPS for Particle Physics and Space Science
2-year PPARC (PPRP) funded programme to develop the underpinning
technology. Started June 2003. Total funding of £300k over 2 years.
5 institutes:
University of Liverpool HEP
University of Glasgow HEP
University of Leicester SS
University of Birmingham SS
CCLRC-RAL with 3 departments: Instrumentation, Space Science and Particle
Physics
4 axes.
1) Pixel architecture: noise, analogue memory, data sparsification
2) Radiation resistance
3) Backthinning: for EUV detection and minimise material for HEP
4) Large area sensor
7
RAL, 16/2/2005
Basic Technology MI-3 consortiumConsortium of 11 institutes http://mi3.shef.ac.uk.
Goal: underpinning technology. University of Sheffield (Department of Electrical and Electronic Engineering) University of Liverpool (Semiconductor Detector Center) University of Liverpool (Lab. For Environmental Gene Regulation) University of Glasgow (Particle Physics Experimental Group) University of Brunel (Imaging for Space and Applications) University College, London (Radiation Physics) Institute for Cancer Research (Royal Marsden Hospital) University of Surrey (Centre for Vision, Speech and Signal Processing) University of York (Applied Electromagnetical and Electron Optics Research Group) MRC Laboratoy of Molecular Biology CCLRC-RAL (EID, PPD, SSTD)
Started 7/04 end 7/08
8
RAL, 16/2/2005
RAL_HEPAPS family.Sensors for Particle Physics
Sensors Year N. pixel Pitch (m)
Epi layer (m)
Technology
HEPAPS1 2001 8*64=0.5K 15 2 0.25 IBM
HEPAPS2 2003 384*192=73K 15 80.25 TSMC CIS(0.35 in pixel)
HEPAPS3 2004 192*192=36K 15 No0.25 TSMC MM-RF(0.35 in pixel)
HEPAPS4 2005 1024*384=300K 15 >10 0.35 AMS OPTO
9
RAL, 16/2/2005
RAL_HEPAPS 2Parametric test
sensor
• 4 pixel types
– 3MOS
– 4MOS
– CPA (charge amp)
– FAPS (10 deep pipeline) R
ow d
ecod
er/c
ontr
ol
3MOSdes. A
3MOSdes. B
3MOSdes. C
3MOSdes. D
3MOSdes. E
3MOSdes. F
4MOSdes. A
4MOSdes. B
4MOSdes. C
4MOSdes. D
4MOSdes. E
4MOSdes. F
CPAdes. A
CPAdes. B
CPAdes. C
CPAdes. D
FAPSdes. A
FAPSdes. B
FAPSdes. C
FAPSdes. D
FAPSdes. E
Columnamplifiers
Column decoder/control
• 3MOS & 4MOS: six different design
each of 64x64 pixels at 64x64, 15m
pitch, 8m epi-layer MIP signal
~600 e-
10
RAL, 16/2/2005
Soft and hard resetRESET
ROW_SELECT
Output
Diode
Reset (or kTC) noise is generally the dominant noise sourceVreset
Hard resetRESET – Vreset > Vth for reset transistor
Noise (ENC in e- rms)
Soft resetRESET ~ Vreset.A factor of ~2 reduction noise < 20 e- rms
Noise (ENC in e- rms)
Measured noise distributions for a 64x64 pixel test structure.
Not corrected for system noise
11
RAL, 16/2/2005
HEPAPS2: Some Clusters
Number of pixels in a “3x3” cluster
Cluster in S/N
Source (Ru106) test results. Test made in Liverpool.
12
RAL, 16/2/2005
HEPAPS2:
Landau
distributions
13
RAL, 16/2/2005
RAL_HEPAPS2 3&4 MOS summary
• All 12 substructures are working.• 2 had initial problems in fabrication, and no time yet to test them with MIPs.• The 3 structures with 4MOS-GAA have S/N too low for efficient use for MIP detection.• All the 7 others display good S/N for MIP detection.• Test beam just finished (results now to be analysed):
– Seed cut determines S/N result– Efficiency, global and as a function of impact point
Type Specs S N S/N
3MOS E 4 diodes 99±1 4.94±0.02 20.1 ±0.3
3MOS C GAA 87±2 4.85±0.02 18.0 ±0.4
3MOS B Diode 1.2x1.2 92±1 3.87±0.01 23.8 ±0.2
3MOS A Diode 3x3 67±1 3.31±0.01 20.3 ±0.3
4MOS C Lower VT 101±2 4.14±0.02 24.4 ±0.4
4MOS B Higher VT 114±2 4.70±0.02 24.2 ±0.4
4MOS A Reference 111±2 4.45±0.02 25.0 ±0.4
8 m epi layer
14
RAL, 16/2/2005
Radiation test. Source results
• Noise seems to
increase slightly with
dose.
• Signal decreases with
dose.
3MOSA 3x3 m2
3MOSB 1.2x1.2 m2
3MOSC GAA
3MOSE 4 diodes
4MOSA Reference
4MOSB Higher VT
4MOSC Lower VT
J. Velthuis (Liv)
15
RAL, 16/2/2005
S/N dependence on impact point.
1015
No rad1014
No rad
4-diode15 m pixel
G. Villani (RAL-PPD)
16
RAL, 16/2/2005
Radiation test. Summary
• Sensors yield reasonable S/N up to 1014 p/cm2 (device simulation confirms)
– No efficiency measurement; need testbeam data• 0.35 m technology in the pixel transistors. Enclosed layout in 3MOS_C• Especially 3MOS_E (4 diodes) looks interesting
– Larger capacitance yields larger noise– Four diodes: less dependence of S/N on impact point– After irradiation remains a larger “sensitive area”
!
-
+
+ J. Velthuis (Liv)
17
RAL, 16/2/2005
Flexible Active Pixel Sensor
10 memory cell per pixel
28 transistors per pixel
20 m pitch
40x40 arrays
Design for the Vertex detector at the International Linear Collider
Pulses LED test
Light pulse
Amplitude
Time
RAL, 16/2/2005
CMOS sensors for the linear collider50 mm
Readout direction
Red line: control electronics (sampling and readout). Minimal space.
Red rectangle: readout electronics (column amplifiers + ADC + sparsifying circuit) Either on same substrate or bump-bonded to sensor substrate
Ladder with 1 (2) sensor(s)Sensor size: 100 mm *13 mm, read out at both sidesNumber of pixels per sensor: 2500 x 650In each pixel: 20 samples
For ILC: sample at 50 s during beam-on periods and store 20 samples in the pixelColumn parallel readout between trains on multiple lines @ 1 MHz a few ms read-out time
50 mm
13 mm
Minimize budget material in the central area
Keep power dissipation evenly spread and low
Keep sensor architecture simple and adaptable to
machine choice
Simplify system design
Guidelines
19
RAL, 16/2/2005
FAPS source test
• Correlated Double Sampling readout (subtract Scell 1)
• Correct remaining common mode and pedestal
• Calculate random noise– Sigma of pedestal and common
mode corrected output• Cluster definition
– Signal >8 seed– Signal >2 next
• Note hit in cell i also present in cell i+1.
• S/Ncell between 14.7±0.4 and 17.0±0.3
Seed 3x3 5x5
J. Velthuis (Liv)
20
RAL, 16/2/2005
FAPS Hit resolution
• Take hits found in cell 2• Reconstruct x and y each
cell using Centre-of-Gravity• Calculate average hit
position• Determine residual position
for each memory cell• Hit resolution approximately
1.3 m• Hit Resolutionspatial
resolution!!!
J. Velthuis (Liv)
21
RAL, 16/2/2005
FAPS efficiency estimate• Find hits in all cells• Plot max S/Npixel in 3x3 area
around expected hit position if hit not found
• Define:
• Clearly, strongly dependent on seed cut. Lowering seed cut to 5 yields inefficiency ranging between 0.08±0.08% and 0.5±0.1%
1)-(i cell seeds#
cut seedmissed#Missed
J. Velthuis (Liv)
22
RAL, 16/2/2005
MAPS plan
Design:
Phase 1: dedicated run with several test structure for FAPS: simulation and test
results to analyse effects of storage cell size, numbers, read-out speed
Phase 2: non-stitched 2 cm x 2 cm uniform sensor
Phase 3: full size, stitched layer 1 sensor
Mechanical mounting: CVD diamond
Readout ASIC. ‘Alignment’ to CCD/ISIS readout cost saving
Explore possibility of sharing technology (ISIS, readout circuit) cost saving.
23
RAL, 16/2/2005
MAPS under LCFI
Groups involved: Liverpool, Glasgow, RAL-EID, RAL-PPD (?)
3 runs to XFAB ~ £200k ( Jazz025 ?)
CVD diamond ~£40k
PCB ~ £20k
DAQ ~ £10k
Design effort (RAL EID): about 3.8 SY. Cost £250k (to be discussed within MI-3)
Technician in Liverpool (50% over 5 years): £75k
Some effort from RA already requested within LCFI
RAs from rolling grants in Liverpool and Glasgow
Total cost: ~£600-700k