a low noise and high sensitivity image sensor with imaging … workshops/2015 worksho… · ·...
TRANSCRIPT
A Low Noise and High Sensitivity Image Sensor
with Imaging and Phase-Difference Detection AF in All Pixels
M.Kobayashi, M.Johnson, Y.Wada, H.Tsuboi, T.Ono, H.Takada, K.Togo,
T.Kishi, H.Takahashi, T.Ichikawa, S.Inoue
Canon Inc., 70-1, Yanagi-cho, Saiwai-ku, Kawasaki-shi, Kanagawa 212-8602, Japan
Phone: +81-3-3758-2111, Fax: +81-44-520-3218
E-mail: [email protected]
Abstract
In this paper, we describe a device structure and
optical design for a CMOS image sensor with the
phase-difference detection photodiodes (PD) for an
autofocus (AF) function. This image sensor has a
pixel separated into two PDs by a PN junction. All
the effective pixels function as both the imaging
and the phase-difference detection AF (PDAF). We
have realized a low dark random noise (=1.8e- at
1PD, 2.5e- at 1pixel) and high sensitivity
(=78,000e-/lx sec at 1green pixel) image sensor with
the imaging and the PDAF functions in all the
effective pixels.
Introduction
Recently, image sensors with the PDAF function at
image plane has been developed [1], [2]. These
realize the AF function by arranging exclusive
pixels to detect the phase-difference. In these image
sensors, a pair of partially light shielded PD is
arranged in a part of pixel array. The focusing speed
is extremely fast in comparison with conventional
contrast detection AF. The sensitivity, however, is
degraded by the light shielding structure and
interpolation processing by neighboring pixels is
necessary for image generation [2]. Therefore, the
number of pixels for AF function is limited to avoid
the deterioration of image quality.
In this work, we developed an image sensor with
the imaging and the PDAF functions in all the
effective pixels without exclusive pixel for AF
function. All the effective pixels have two PDs each
to detect phase-difference in one pixel without
partially light shielding structure. Therefore, one
pixel works as one AF point, and the sum of two PD
outputs equals one pixel output.
Fig. 1: Principle of the PDAF function
Principle Fig. 1 shows the principle of the PDAF function.
Exit pupil of an optical lens and a PD of the image
sensor are in optically conjugate relation by on-chip
micro-lens (ML). Therefore, each pixel by separating
into two PDs has the pupil split function for the
PDAF function. A light flux which passes through a
right half of the optical lens is led to the PD-A (the
PD at left si
of the optica
side).
Defocusing
between th
provided fro
image-B (th
The directi
image-B is
and the back
These mea
lens is drive
defocusing a
Fi
Fig. 2 show
pixel consist
ML and on
consists of o
transistors
signal line (
eight transi
is symmetri
60frames pe
driven simu
Fig. 3 show
The PD-A
junction. T
consist of t
using the
area is m
boundary
ide) and whic
al lens is led t
g amount is
he peak of
om PD-A gr
he image prov
ion of the pe
opposite bet
k focus state
an that in the
en to a just f
amount and t
ig. 2: Schema
Pixel A
ws the schem
ts of two sub
ne color-filter
one PD, one
for one sign
CSL), thus o
istors and tw
ically arrang
er second re
ultaneously.
ws the cross s
and the PD
The PN jun
the insulatin
PN junction
minimized, li
is reduced
ch passes thr
to the PD-B
s calculated
the image
roup) and th
vided from PD
eak shift of th
tween the fr
e.
e camera sys
focus instant
the direction
atic view of 2x
Architecture
matic view of
b-pixels, and
r (CF) each.
floating diffu
nal output a
one pixel cons
wo CSLs. A p
ged left and r
adout speed
section diagr
D-B are sepa
nction separa
ng materials
n separation
ight reflecti
and defec
rough a left h
(the PD at ri
from dista
e-A (the im
he peak of
D-B group).
he image-A a
ront focus st
stem the opt
tly by calcula
n [3 .
x2 pixels
e
2x2 pixels. O
has one on-c
. One sub-pi
usion (FD), f
and one colu
sists of two P
pair of sub-pi
right. To achi
, four CSLs
ram of the pix
arated by a
ation does
(e.g. SiO2).
n, non-sensit
ion at Si/Si
cts at Si/Si
half
ight
ance
age
the
and
tate
tical
ated
One
chip
ixel
four
umn
PDs,
ixel
ieve
are
ixel.
PN
not
By
tive
iO2
iO2
inte
shr
and
In
p-ty
low
pin
p-ty
loss
ligh
nois
W
tim
tran
the
Fi
dep
per
ima
sen
the
pos
the
Th
dep
PD
cas
sim
high
cas
sep
erface is decr
rinkage at Si
d defects incr
n addition, a
ype region f
wer than tha
nned PD. Th
ype impurity
s of the incid
ht is divided i
se and high s
Fig. 3: Cro
We performed
me domain (
nsient analy
curvature of
g.4 shows
pendence of
rpendicular g
age sensor su
nsor structure
e light inte
sition of the i
e ML curvatu
he figure als
pendence of
-A, PD-B an
e (b) (the rad
mulated QE o
her than the
e (c) (RC=5.3
parated PDs,
reased. It is
i/SiO2 interfa
rease [4], [5].
an impurity
for the PN
at of the surf
herefore, reco
y separation
dent light is m
into PD-A an
sensitivity im
oss section di
Optical D
a three-dime
(FDTD) opt
ysis device s
f the on-chip
the simu
f the light
green inciden
urface, supe
e diagram. A
ensity is st
incident light
ure.
so shows th
the quantu
nd the sum
dius of curva
of a two PD
e case (a) (RC
3 m) 55.1%. B
the sensitiv
reported tha
ace causes d
y concentrat
junction se
face region f
ombination
n region is lo
minimized. T
nd PD-B, as a
mage sensor i
iagram of the
Design
ensional finit
tical simulat
imulation to
ML.
ulated ML
intensity p
nt light ( =55
erimposed on
A dark color
trong. The
ht changes de
he incident
um efficienci
of the two P
ature (RC) is
s summation
C=3.3 m) 51.
By adding sig
vity is equiva
at band-gap
dark current
tion of the
eparation is
forming the
rate in the
ow, and the
The incident
a result, low
is realized.
e pixel
te difference
tion and a
o determine
curvature
profiles for
50nm) to the
n the image
means that
condensing
epending on
light angle
ies (QE) of
PDs. In the
4.3 m), the
n is 59.5%.,
.6% and the
gnals of two
alent to that
of one PD st
The bottom
angle depen
in the posit
PD-A in the
below the ra
the case (b
range is wid
case (c) 24
image-A an
range of th
performanc
tructure.
mmost figur
ndence of the
tive angle re
e negative an
atio of 0.2 is
b), the angle
der than the
degrees. Th
nd the imag
he incident
e is improved
Fig. 4
Incide
re shows the
e QE ratio of
egion and ra
ngle region. T
compared in
e range is 2
e case (a) 21 d
hat wider ra
ge-B can sep
light angle
d. Similarly,
4: Simulated
ent light angl
e incident li
f PD-A to PD
atio of PD-B
The angle ran
n three cases
25 degrees. T
degrees and
ange means
parate at wi
e, and the
the angle ran
ML curvatur
le dependenc
ight
D-B
B to
nge
. In
The
the
the
ider
AF
nge
abo
the
is n
cas
deg
ima
ape
AF
ML
sim
sha
re dependenc
ce of the QE a
ove the ratio o
case (b), the
narrower tha
e of (c) 15 d
grees means
age-B is bet
erture). To re
performance
L with the
mulation of t
ape of the ML
ce of the light
and the ratio
of 0.2 near 0
e angle range
an the case o
degrees. The
the separat
tter at large
ealize both hi
e in the desi
e PDAF fu
the height,
L is required.
t intensity pr
o of PD-A and
degrees is co
e is 9 degrees
of (a) 10 degr
narrower ra
tion of image
er F-number
high sensitivit
ign of two P
unction, the
the curvatu
rofiles,
d PD-B
ompared. In
s. The range
rees and the
ange near 0
e-A and the
r (=smaller
ty and high
PDs and one
e accurate
ure and the
Table 1 su
performanc
process is
Optical form
6.4 m. Nu
points are b
18.5M. Full
two PDs). S
source with
random noi
pixel (@gain
pixel (@60
60frames pe
Fig. 5 show
sensor. Chip
High grad
achieved by
image sens
detection AF
The autho
T.Kato,
A.Takabaya
of Canon In
References
[1] S.Uchiya
ITE Tec
[2] H.Endo,
The jou
[3] I.Onuki,
The jou
[4] K.Takeu
Applied
[5] K. Itonag
Electron
Sum
ummarizes
es of this
0.18 m 1Po
mat is super 3
umber of effe
both 9.2M. N
l well capacit
Sensitivity is
h IR-cut-filte
ise is 1.8e- a
n=32, RT). D
, one secon
er second wit
ws the packa
p size is 29.9m
de and high
y the low n
sor with ima
F in all pixels
Acknow
ors would li
J.Iwata,
ashi, I.Onuki
nc. for their co
s
ama, �Superi
chnical Repor
��Phase Det
urnal of the In
�Digital Cam
urnal of the In
uchi, et al., �S
d physics lette
ga, et al., "Ex
n Devices Me
mmary
the specifica
image sens
oly 4Metal C
35mm. Pixel
ective pixels
Number of e
ty is 40,000e
78,000e-/lx s
er, one gree
at one PD an
Dark current
d). Maximum
th full pixel r
age photograp
mm x 21.6mm
h speed ima
noise and h
aging and p
s.
wledgement
ike to thank
Y.Arishim
i and Y.Hara
ontribution to
iority of Imag
rt, 36.38 (201
tection Pixel B
nstitute of Im
mera with Ph
nstitute of Im
i band-gap sh
ers, 61.21, (1
xtremely-Low
eeting (IEDM
ations and
sor. Fabricat
CMOS proce
size is 6.4 m
and the PD
effective PDs
e- (@one pixe
s (@2,856K li
en pixel). D
nd 2.5e- at o
t is 50e- at o
m frame rate
readout.
ph of this im
m.
aging has be
high sensitiv
phase-differe
t
k K.Kawaba
ma, A.Ok
ada the mem
o this work.
ge Plane Pha
12): 17.
Built-in Imag
mage Informa
hase Detectio
mage Informa
hrinkage cau
1992): 2566-2
w-Noise CMO
M), 2011 IEE
the
tion
ess.
m x
DAF
s is
el =
ight
ark
one
one
e is
age
een
vity
ence
ata,
kita,
mber
F
ase Detection
ge Sensor� to
ation and Tel
on Image Sen
ation and Tel
used by local
568.
OS Image Sen
E Internatio
Table 1: S
perform
Fig. 5: Packag
n AF�,
o Realize Hig
levision Engi
nsor�,
levision Engi
strain at Si/S
nsor with Hi
onal. IEEE, 2
Summarized
mances of thi
ge photograp
gh Speed Aut
ineers, 65.3 (2
ineers, 68.3 (2
SiO2 edge�,
igh Saturatio
011.
specification
is image sens
ph of this ima
to Focus�,
2011): 290-29
2014): 203-20
on Capacity",
ns and
sor
age sensor
92.
07.
,