ESA ITT AO/1-3970/02/NL/ECemits.sso.esa.int/emits-doc/ESTEC/AO6131hybridlpresfinal.pdfESA ITT...
Transcript of ESA ITT AO/1-3970/02/NL/ECemits.sso.esa.int/emits-doc/ESTEC/AO6131hybridlpresfinal.pdfESA ITT...
-
Radiometric performance enhancement of
Active Pixel Sensors
ESA ITT AO/1-3970/02/NL/EC
Koen
De Munck, Deniz
Sabuncuoglu
Tezcan, Kiki Minoglou,
Piet De Moor, Jean Roggen, Chris Van Hoof
IMEC
Jan Bogaerts
Fillfactory/Cypress
Iacopo
Ficai
Veltroni, Cinzia
Toccafondi
Galileo Avionica
-
Piet De Moor imec 2008 2
Outline
Introduction
CMOS APS/readout design
Technological developments
Characterization
Conclusions & outlook
-
Piet De Moor imec 2008 3
ITT by the European Space Agency in 2002
Aim:
snapshot shutter CMOS APS demonstrator for high-performance spaceborne
imaging (BI CCD replacement)
Application:
Earth observation
Hyperspectral
imaging
Partners:
FillFactory/Cypress: design
IMEC: technology development
Galileo Avionica: radiometric characterization
Introduction: Aim & collaboration
-
Piet De Moor imec 2008 4
Array size: 1024x1024
Pixel pitch: 22.5 micron
Filling Factor: > 0.95
Spectral range: 0.39
1 micron
Quantum efficiency:
> 80% in 0.39
0.80 micron
> 50% in 0.8
0.9 micron
Dark current < 1500 e-/(pix.s) @ -15C
Read noise < 50 e-
Full well capacity: > 8 105
e-
Non-linearity: < 1% p-p
Cross-talk < 10-4
Snapshot, windowing
Introduction: Requirements
backside illuminated imager
CMOS active pixel sensor
-
Piet De Moor imec 2008 5
Introduction: Approach: monolithic & hybridMonolithic approach
backside thinned CMOS APS mounted on MCM substrate
using Au stud bumps MCM substrate
backside thinned diode arrayinterconnected pixel-by pixel
to CMOS ROICby In bumps
Hybrid approach
ROIC
CMOS APS (mono) = ROIC (hybrid):
designed by FF/Cypress, manufactured in XFAB 0.35 um technology
Diode array: manufactured in IMEC
-
Piet De Moor imec 2008 6
Outline
Introduction:
CMOS APS/readout design:
Imager and pixel architecture
Readout modes
Output stage
Technological developments
Characterization
Conclusions & outlook
-
Piet De Moor imec 2008 7
Design: ROIC architecture
synchronous pipelined shutter
22.5 m pixel pitch
stitched design:
512 x 512 pixels stitch blocks
up to 2048 x 2048 pixels
pseudo-differential output
per 256 columns
20 Mpixels/s
per output
SPI interface for upload:
addressing, gain & offset, NDR, non-linear amplifier,
etc.
radiation tolerant design
-
Piet De Moor imec 2008 8
20
48
x 2
04
8
8 o
utp
uts
Design: Read-out speed
Frame rate is determined by output speed and window size
Multiple window read-out limited by row overhead time
frame time (ms)
Frame rate (/s)
Full frame 36.7 27.3
Horizontal band of 150 lines 2.87 3484
10 windows of 30x300 pixels 2.26 442
1 row read-out 0.0178 56180
Full frame 18.43 54.3
Horizontal band of 150 lines 2.07 348.4
10 windows of 30x300 pixels 2.83 354
1 row read-out 0.0178 5618010
24
x 1
02
4
4 o
utp
uts
-
Piet De Moor imec 2008 9
Design: Pixel architecture
monolithic hybrid
in-pixel storage capacitance (fF) 350 350
full well charge (x 1000 electrons) (FWC)
> 200 > 950
read noise (electrons) < 38 < 180
dynamic range (FWC/dark noise) (dB)
75 75
maximum SNR (dB) 53 60
single pixel
sample capacitors
pad for hybridization
photodiode
Cph
determines
charge handling capacity (FWC) and conversion gain
C1,2,3 determine read noise and dynamic range
-
Piet De Moor imec 2008 10
Design: Normal readout mode
synchronous shutter:all pixels integrate in parallel
pipelined:readout while integrate
correlated double sampling
In both cases:
C1 and C3: reset level
C2: signal level
Tint < Tread
Tint > Tread
-
Piet De Moor imec 2008 11
Design: Optimized readout modes I
Non-destructive readout (NDR):
Using the three in-pixel capacitors
Read multiple times without resetting
Advantages: low noise, high dynamic range
-
Piet De Moor imec 2008 12
Design: Optimized readout modes II
Line by line variable integration time:
Integration time selectable per line
Application:
Matching of useful/optimal swing to irradiance
i.e. observation of both bright and dark areas
Reference: J. Bogaerts
et al.: 2005 IEEE Workshop on CCD and Advanced Image Sensors, Nagano, Japan, June 2005
-
Piet De Moor imec 2008 13
Design: Output stage
Non-linear compensation:
Provides inverse response of typical light-to-voltage signal
Programmable settings (knee points offset and slopes)
Gain stage:
16 different gain settings, scaling with 1.1x
Output driver:
20 pF at 20 Mpix/s
No ADCs on chip
-
Piet De Moor imec 2008 14
Outline
Introduction:
CMOS APS/readout design:
Technological developments
Process flow
Thin wafer handling
Trenches
Backside surface treatment & anti reflection coating
Results
Characterization
Conclusions & outlook
-
Piet De Moor imec 2008 15
Technological developments: Graded epi growth
growth of of
50 micron thick epitaxial Si with graded doping
aim: enhanced charge collection due to built-in electrical field
backside
frontside
built-in electric field 1x1016/cm3
1x1014/cm3
3x1015/cm3
3x1014/cm3
1x1015/cm3
start wafer (highly doped)
higher doping
lower doping
1x1016/cm3
1x1014/cm3
3x1015/cm3
3x1014/cm3
1x1015/cm3
start wafer (highly doped)
higher doping
lower doping
-
Piet De Moor imec 2008 16
Technological developments: Imager processing
ROIC design: FillFactory/Cypress
Hybrid diode array design: IMEC
Different imager sizes possible through stitching
Different imager sizes on a 200 mm wafer
1x 20482
16x 10242
20x 5122
on reticle
on wafer
512 x 512 1024 x 1024 2048 x 2048
-
Piet De Moor imec 2008 17
Technological developments: Imager/ROIC processing
Production:
@ XFAB
0.35 m technology
on 200 mm wafers
-
Piet De Moor imec 2008 18
Technological developments: Hybrid diode array processing
Process in CMOS Pilot line IMEC (0.13 m technology baseline) on 200 mm wafers
n-
well implantation
n+ contacts
p+ substrate contacts
W plugs
Al metallization
bulk Si
graded doping epi
n-well p+ contact
-
Piet De Moor imec 2008 19
Technological developments: Trenches for reduced cross-talk
Extra processing steps on hybrid diode arrays:
Trench etching using DRIE, doped poly-Si deposition
Aim: reduce electrical cross-talk
Top view
Cross-section of filled trenches
Cross-section of unfilled trenches
width :1.2 m depth: 50 m
Doped poly-Si filled trenches
np Built-in electric fields
22.5 m
>1e19 at/cm3
-
Piet De Moor imec 2008 20
Technological developments: Imager post-processing: general approach
Challenge: thin wafers are fragile and flexible
Solution:
Use of temporary carriers (Si, glass)
Use of temporary glues
Allows use of standard equipment
50 micron thin 200 mm Si wafercarrier wafer
thin device wafertemporary glue layer
-
Piet De Moor imec 2008 21
Technological developments: Imager post-processing: process flow 1
Process:
Temporary glue layer deposition
Wafer to wafer bonding
Challenges:
Uniform voidless wafer level bonding
Results otherwise in wafer damage during grinding
device wafer
carrier wafer
glue layer
-
Piet De Moor imec 2008 22
Technological developments: Imager post-processing: process flow 2
Ground 300 mm
wafer
False color image:
TTV = 0.8 m
Ground 300 mm
wafer
False color image:
TTV = 0.8 m
Process:
Wafer thinning on carrier to a final thickness of ~35 micron
3 steps: grinding, dry RIE etch, wet etch
Challenges:
Minimized damage at backside, resulting in bad blue response of imager
Obtaining total thickness variation of < 2 micron on wafer level
Measurement of device wafer thickness on carrier
-
Piet De Moor imec 2008 23
Technological developments: Imager post-processing: process flow 3
Process at backside of the wafer:
Backside surface shield in order to repel carriers from surface states:
B implantation at low energy
Laser annealing for low temperature activation
Aluminum shield around imager
Challenge:
Backside surface shield must be very thin in order not to block blue light
Through carrier alignment front -
back
backside of imager with Al shield
-
Piet De Moor imec 2008 24
Technological developments: Imager post-processing: process flow 4
Challenge:
Process thin wafer at the (buried) front side
Solution= thin wafer flip:
Second glue layer (with different release properties) deposition
Second carrier wafer bonding
Release first carrier
-
Piet De Moor imec 2008 25
Technological developments: Imager post-processing: process flow 5 (hybrid only)
Processing:
Under bump metallization (UBM) deposition & patterning
In deposition and patterning
Dicing (including carrier wafer)
Challenge:
In bump uniformity & yield
UBM
In b
um
ps
22.5 um
-
Piet De Moor imec 2008 26
Technological developments: Imager post-processing: process flow 6 for monolithic
Monolithic integration:
Au stud bumping
Au bumps deposited on MCM substrate
Au thermo-compression on Al bonding pads of imager using flip-
chip bonder
Challenge:
Flip-chip of thin die
Solution:
Use carrier die + release
MCM substrate
imager
MCM substrate
imager
carrier
carrier
Au stud bump
-
Piet De Moor imec 2008 27
Technological developments: MCM wafer process
1 metal layer interconnect between landing pad Au stud bump and bonding pad
-
Piet De Moor imec 2008 28
Technological developments: Imager post-processing: process flow 6 for hybrid
Hybrid integration:
UBM & In processed on both diode array and ROIC side
Bumping using flip-chip bonder
Carrier die release
ROIC
carrier
diode array
ROIC
carrier
diode array
diode array
RO
IC
-
Piet De Moor imec 2008 29
Technological developments: Imager post-processing: process flow 7
Mounting on COB board
Using wire bonding
ARC deposition
ZnS
+ MgF2
Using shadow mask
-
Piet De Moor imec 2008 30
Technological developments: Results: monolithic imagers
Monolithic imagers:
1024 x 1024
40m
MCM
imager 40m
MCM
imager 40m
MCM
imager
-
Piet De Moor imec 2008 31
Technological developments: Results: hybrid imagers
50 um
hybrid diode array
ROIC
50 um
hybrid diode array
ROIC
ROIC
hybrid diode array
Hybrid imagers:
512x512
1024x1024
-
Piet De Moor imec 2008 32
Technological developments: Results: special devices
1024x1024 trenched hybrid imager
Unthinned
2kx2k read-out
-
Piet De Moor imec 2008 33
Outline
Introduction
CMOS APS/readout design
Technological developments
Characterization:
Functionality & yield
Response & non-linearity
Dark current & read noise
PRNU & Quantum efficiency
Cross talk
Radiation hardness
Conclusions & outlook
-
Piet De Moor imec 2008 34
Characterization: functionality check
2048x2048 direct imaging of
hexagonal nut
2048 x 2048:
unthinned
read-out
front side illumination
high yield:
2 bad columns, 2 partial rows
> 99.8 %
1048 x 1048:
Hybrid thinned device
Monolithic thinned device
Trenched Hybrid thinned device
512 x 512:
Test devices: hybrid, thinned & unthinned
Hyb
rid 1
024x1
024
-
Piet De Moor imec 2008 35
0.01
0.1
1
10
100C
umulative pixelcount (%
)
1.00.80.60.40.20.0Intensity (normalized scale)
80x103
60
40
20
0
#pix
els
(arb
. sca
le)
Characterization: Pixel yield analysis of Hybrid
Hybrid thinned 1024x1024 imager
Yield in terms of light sensitivity: 99.93% < Y < 99.99%
99.93%99.99%
Interconnect yield
Pixels with nominal dark
current
Analysis on differential image (Bright
Dark)
-
Piet De Moor imec 2008 36
Characterization: Pixel yield analysis of Monolithic
Monolithic thinned 1024x1024 imager
Yield in terms of light sensitivity > 99.999 %
When removing 10 faulty lines (~ 1 %)
10-1100101102103104105
#pix
els
(arb
. sca
le)
0.300.200.100.00Intensity (normalized scale)
0.0001
0.001
0.01
0.1
1
10
100 Cumulative pixelcount(%
)
-
Piet De Moor imec 2008 37
Characterization: Response conversion factor
Conversion factor CVF:
Dependent on gain and NLA settings
Mono: 6 -
12 uV/e-
Hybrid: 1.5
3 uV/e-
hybrid
Full well capacity FWC:
Mono: 1.8 105
e-
Hybrid: 7.57 105
e-
0.00
10.00
20.00
30.00
40.00
50.00
60.00
0 2 4 6 8 10 12 14 16
Gain settingC
onve
rsio
n ga
in [u
V/e]
1.05
1.07
1.09
1.11
1.13
1.15
1.17
Gai
n (x
+) /
Gai
n (x
)
mono
-
Piet De Moor imec 2008 38
Characterization: Non-linearity of response
Non-linearity of monolithic device with non-linear amplifier OFF:
1% peak-peak in the range 20%
64% of full well capacity
mono
% d
evia
tion
from
line
ar fi
t
-
Piet De Moor imec 2008 39
Characterization: Non-linearity of response
Non-linearity of monolithic device with non-linear amplifier ON, using optimized settings:
0.5% peak-peak in the range 8%
87% of full well capacity
Conclusion: excellent linearity using correction
mono
% FW
% d
evia
tion
from
line
ar fi
t
-
Piet De Moor imec 2008 40
Characterization: Dark current
Dark current at -15C
Mono: 330 e-/pix.s, corresponding to 0.01 nA/cm2
Hybrid: 736 e-/pix.s, corresponding to 0.023 nA/cm2
Temperature dependence: *2 every 8-9C
Dark current non-uniformity: < 20
50 %
Temperature MonoDark current (e-/pix.s)
HybridDark current (e-/pix.s)
-20 218 875
-15 330 736
-10 484 1696
0 1404 3838
10 2552 9838
20 6766 19690
30 13513 45582
mono
-
Piet De Moor imec 2008 41
Characterization: Read noise
Mono:
@ 25C: 88 e-
(715 uV)
@ -15C: 80 e-
(642 uV)
Hybrid @25C:
142
264 e-
(highest gain, unity gain)
Corresponding to 794, 249 uV
Remark on noise numbers:
Including system noise !
= upper limit
hybrid
hybrid
-
Piet De Moor imec 2008 42
Characterization: Pixel response non-uniformity (PRNU)
Mono:
Zone A: 15.9 %
Zone B: 5.01 %
Zone C: 8.24 %
Zone A: 9.13 %
High frequency ~ 4.25 %
Hybrid:
Zone A: 11.8 %
Zone B: 8.49 %
Zone C: 8.23 %
Zone A: 8.54 %
High frequency ~ 7.4 %
High PRNU most probably related to residues of temporary glue on device backside
will be addressed in near future
hybrid
mono flat field
-
Piet De Moor imec 2008 43
Characterization: Quantum efficiency: impact of backside passivation
QE measured without anti-refection coating
After backside passivation: increased blue response !
Effective non-sensitive thickness layer ~ 10
15 nm
0
10
20
30
40
50
60
70
80
300 400 500 600 700 800 900 1000 1100
Wavelength [nm]
Qua
ntum
effi
cien
cy [%
]
measured 22 m thick
15 nm
20 nm
100 nm
10 nm
50 nm
0 nm
dead layer thickness
before laser annealing
5 nm
2 nm
-
Piet De Moor imec 2008 44
Characterization: Anti Reflective Coating (ARC)
Optimized broad band ARC
Reflectivity
-
Piet De Moor imec 2008 45
Characterization: Quantum efficiency: backside passivation + ARC
Excellent broad band QE > 80% from 400 nm
850 nm
Similar results for both mono and hybrid devices
0
10
20
30
40
50
60
70
80
90
100
250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000 1050 1100
Wavelength [nm]
Qua
ntum
effi
cien
cy [%
]
measurement
simulation: plain Si0 nm
10 nm15 nm
simulation: ARC limited
0 nm15 nm simulation: ARC + implant
0 nm10 nm
15 nm
dead layer thickness
QE > 80 %
-
Piet De Moor imec 2008 46
Characterization: Quantum efficiency: degradation of ARC
Reduction of QE as a function of time
= degradation of ARC probably due to humidity absorption
will be addressed in near future
0
10
20
30
40
50
60
70
80
90
100
250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000 1050 1100
Wavelength [nm]
Qua
ntum
effi
cien
cy [%
]
back_1k_dev4 with ARC (may 9)
back_1k_dev4 with ARC (june 19)
back_1k_dev4 no ARC (april 4)
-
Piet De Moor imec 2008 47
Characterization: Quantum efficiency: trenched hybrid imager
QE of trenched hybrid imager is low after backside passivation
This is probably due to:
reduced fill factor
recombination at trench sidewalls
No trenches
Trenched Trenched compensated for fill factor loss No ARC
-
Piet De Moor imec 2008 48
Characterization: Cross - talk
Knife edge measurement + MTF calculation
Good MTF for trenched hybrid imager: 0.5 at Nyquist
0
10
20
30
40
50
60
70
80
90
100
0 10 20 30 40 50 60 70 80 90 100
Knife edge position [um]
Rel
ativ
e pi
xel r
espo
nse
[%]
1k trenched - 550 nm1k trenched - 650 nm1k trenched - 450 nm512 no trench - 550 nm512 no trench - 650 nm512 no trench - 450 nmideal curvepixel border
effect of trenches
0.00 0.02 0.04 0.06 0.08 0.100.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
MTF
spatial Frequency lp/um
ampl
itude
-
Piet De Moor imec 2008 49
0.000 0 .0 05 0.01 0 0 .015 0.0 200.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Frequency (lp/um)
MT
FCharacterization: Cross - talk
Point source illumination + MTF calculation
Bad MTF for monolithic device:
Close to 0 at Nyquist
mono
60x Microscope objective
x-y-z translation stage
Neutral density filter assembly
Low power beam
HeNe lasercollimator
60x Microscope objective
x-y-z translation stage
Neutral density filter assembly
Low power beam
HeNe laserHeNe lasercollimatorcollimator
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
-
Piet De Moor imec 2008 50
Characterization: Cross - talk
Monolithic device:
Signal leakage to neigbouring
pixels = 88 %
Caused by diffusion of carriers
Trenched hybrid device:
Signal leakage to neigbouring
pixels = 0 %
Diffusion of carriers to neigbouring
pixels is
blocked
12
3
S1
S2
S3
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1
2
3
4
5
S1S2
S3S4
S5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
-
Piet De Moor imec 2008 51
Characterization: Radiation tests
Proton irradiation
5 x 1010
60 MeV
protons/cm2
Additional 60Cogamma irradiation
50 krad(Si)
Dark current measured before and after irradiation:
512x512 hybrid device:
Increase of dark current by factor 2.5 after proton irradiations
Additional increase of dark current by a factor of 17 after gamma radiation
512x512 mono device, unthinned, front side illumination
No increase of dark current after proton irradiation, # dark current spikes increases
Increase of dark current by a factor of 4 after gamma radiation
0.001
0.01
0.1
1
10
100
0 200 400 600 800 1000 1200 1400 1600
Dark current [mV/s]
Cum
ulat
ive
freq
uenc
y [%
]
device 2 pre
device 2 post 5x10^10protons/cm^2device 2 post 50 krad(Si)
device 4 pre
device 4 post 5x10^10protons/cm^2device 4 post 50 krad(Si)
device 3 pre
-
Piet De Moor imec 2008 52
Outline
Introduction
CMOS APS/readout design
Technological developments
Characterization
Conclusions & outlook
Conclusions
Outlook
-
Piet De Moor imec 2008 53
Conclusions
Thinned backside illuminated imagers realized:
Hybridized and monolithic
CMOS APS: synchronous pipelined shutter with true CDS
3D process technology for thin wafer handling and processing:
Use of temporary carriers and glues
200 mm thin wafer processing on carrier with standard equipment
Performance enhancing concepts implemented:
Graded EPI lower cross-talk and improved QE
Poly-Si filled pixel separating trenches zero cross-talk (but QE )
Results:
Excellent broadband QE (but AR ageing )
Zero cross-talk for trenched hybrid device
-
Piet De Moor imec 2008 54
Outlook
Further investigations planned
Low cross talk while maintaining high QE
Stabilize ARC
Improve PRNU
Process optimizations (i.e. monolithic imager yield)
CCN on this contract under discussion
Performance enhancement using 3D integration technology: through Si vias
detection layer
analog ROIC
digital signal processing
memory
output drivers
-
Piet De Moor imec 2008 55
Radiometric performance enhancement ofActive Pixel SensorsOutlineIntroduction:Aim & collaborationIntroduction:RequirementsIntroduction:Approach: monolithic & hybridOutlineDesign:ROIC architectureDesign:Read-out speedDesign:Pixel architectureDesign:Normal readout modeDesign:Optimized readout modes IDesign:Optimized readout modes IIDesign:Output stageOutlineTechnological developments:Graded epi growthTechnological developments:Imager processingTechnological developments:Imager/ROIC processingTechnological developments:Hybrid diode array processingTechnological developments:Trenches for reduced cross-talkTechnological developments:Imager post-processing: general approachTechnological developments:Imager post-processing: process flow 1Technological developments:Imager post-processing: process flow 2Technological developments:Imager post-processing: process flow 3Technological developments:Imager post-processing: process flow 4Technological developments:Imager post-processing: process flow 5 (hybrid only)Technological developments:Imager post-processing: process flow 6 for monolithicTechnological developments:MCM wafer processTechnological developments:Imager post-processing: process flow 6 for hybridTechnological developments:Imager post-processing: process flow 7Technological developments:Results: monolithic imagersTechnological developments:Results: hybrid imagersTechnological developments:Results: special devicesOutlineCharacterization:functionality checkCharacterization:Pixel yield analysis of HybridCharacterization:Pixel yield analysis of MonolithicCharacterization:Response conversion factorCharacterization:Non-linearity of responseCharacterization:Non-linearity of responseCharacterization:Dark currentCharacterization:Read noiseCharacterization:Pixel response non-uniformity (PRNU) Characterization:Quantum efficiency: impact of backside passivationCharacterization:Anti Reflective Coating (ARC)Characterization:Quantum efficiency: backside passivation + ARCCharacterization:Quantum efficiency: degradation of ARCCharacterization:Quantum efficiency: trenched hybrid imagerCharacterization:Cross - talkCharacterization:Cross - talkCharacterization:Cross - talkCharacterization:Radiation testsOutlineConclusionsOutlookSlide Number 55