How Small Should Pixel Size Be?
Ting Chen1, Peter B. Catrysse1, Abbas El Gamal1 and Brian A. Wandell2
1Information Systems Laboratory
Department of Electrical Engineering, Stanford University, CA 94305, USA
2Department of Psychology, Stanford University, CA 94305, USA
EI 2000, 3965-53 1
Pixel Design
• Pixel design is crucial in image sensor design
– Selection of pixel architecture
– Selection of photodetector type
– Selection of pixel layout and size
• Pixel size tradeoff
– Small pixel size/large pixel count
improves spatial resolution and MTF
– Large pixel size improves DR and SNR
=⇒ an optimal pixel size may exist
EI 2000, 3965-53 2
Pixel Size Selection
• Difficult to determine the optimal pixel size
analytically, depends on
– Sensor parameters
– Imaging optics
– Human perception of image quality
• We describe a methodology using
– Camera simulator (Catrysse et al. 1999)
– Image quality metric S-CIELAB (Zhang et al.
1997)
EI 2000, 3965-53 3
Outline
• APS pixel circuit and layout
• Pixel size tradeoff
– DR and SNR
– Spatial Resolution and System MTF
• Methodology for determining optimal pixel size
• Simulation results
• Conclusion
EI 2000, 3965-53 4
APS Pixel Circuit and Layout Topology
vdd
iph + idc
Cpd
Co
M1
M2
M3
M4
Reset
Word
Bias
IN
Bitline OUTColumn and Chip
Level Circuits
EI 2000, 3965-53 5
DR and SNR versus Pixel Size
5 6 7 8 9 10 11 12 13 14 1535
40
45
50
55
60
65
70
DR
SNR @ 20% Well Capacity
Pixel size (µm)
DR
and
SN
R (
dB)
EI 2000, 3965-53 6
Spatial Resolution and System MTF versus Pixel Size
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
6µm8µm
10µm12µm
Normalized spatial frequency
MT
F
EI 2000, 3965-53 7
Methodology for Determining Optimal Pixel Size
• Given :
– Process information
– Fixed die size
– Imaging optics parameters
– Scene luminance range
– Highest spatial frequency in the scene
– Maximum exposure time
EI 2000, 3965-53 8
Procedure for Determining Optimal Pixel Size
• Vary pixel size (pixel count) for the given die size
• For each pixel size
– Use the camera simulator with a synthetic
Contrast Sensitivity Function (CSF) scene to
estimate the resulting image
– Determine rendered image quality in
S-CIELAB
• Select optimal pixel size to achieve highest image
quality
EI 2000, 3965-53 9
Synthetic Contrast Sensitivity Function Scene
Spatial frequency (lp/mm)
Con
tras
t
5 10 15 20 25 30
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
EI 2000, 3965-53 10
Simulation Flow
EI 2000, 3965-53 11
Example : 0.35µ CMOS Process
pixel size : 8µm
Lscene : 25 - 1000 cd/m2
tmax : 100 ms
fmax : 33 lp/mm
optics f/# : 1.2
5 10 15 20 25 30
0.2
0.4
0.6
0.8
1
5 10 15 20 25 30
0.2
0.4
0.6
0.8
1
5 10 15 20 25 30
0.2
0.4
0.6
0.8
1
5 10 15 20 25 30
0.2
0.4
0.6
0.8
1
Spatial frequency (lp/mm)Spatial frequency (lp/mm)
Spatial frequency (lp/mm)Spatial frequency (lp/mm)
Contr
ast
Contr
ast
Contr
ast
Contr
ast
Perfect Image Camera Output Image
∆E Error Map Iso-∆E Curve
1
2
3
5
EI 2000, 3965-53 12
Determining the Optimal Pixel Size
• 0.35µ process
• fixed die size
• variable pixel size : 5.3 - 15 µm
• scene luminance range : 25 - 1000 cd/m2
• maximum integration time : 100 ms
• highest spatial frequency in the scene : 33 lp/mm
• examine iso-∆E curves for different pixel sizes
• imaging optics f/# : 1.2
EI 2000, 3965-53 13
Iso-∆E = 3 curves for different pixel sizes
5 10 15 20 25 30
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
5.3µm
6µm
8µm
15µm
Spatial frequency (lp/mm)
Contr
ast
EI 2000, 3965-53 14
Average ∆E versus pixel size
5 6 7 8 9 10 11 12 13 14 151.2
1.4
1.6
1.8
2
2.2
2.4
2.6
Pixel size (µm)
Ave
rage
∆E
EI 2000, 3965-53 15
Effect of Technology Scaling on Pixel Size
2 3 4 5 6 7 8 9 10 11 12 13 14 151
1.5
2
2.5
3
3.5
4
4.5
5
5.5
0.35µm
0.25µm
0.18µm
Pixel size (µm)
Ave
rage
∆E
EI 2000, 3965-53 16
Optimal pixel size versus technology
0.1 0.15 0.2 0.25 0.3 0.35 0.40
1
2
3
4
5
6
7
8
Linear Scaling
Simulated
Technology (µm)
Opt
imal
pix
el s
ize
(µm
)
EI 2000, 3965-53 17
Conclusion
• Proposed a methodology using a camera simulator,
synthetic CSF scenes, and S-CIELAB for selecting
the optimal pixel size
• Applied the methodology to photodiode APS
implemented in CMOS technologies down to 0.18µ
• For a 0.35µ process, the optimal pixel size is found
to be around 6.5µm with fill factor 30%
• The optimal pixel size scales with technology,
albeit at slower rate than the technology
EI 2000, 3965-53 18