CMOS Image Sensor Pixel Design and Optimization

1 1 No Security: Public Information Defining the Future of Digital Imaging February 1st 2017 Boyd Fowler cmos Image Sensor Pixel Design and Optimization 2 No Security: Public Information Defining the Future of Digital Imaging * Outline Introduction Photodetectors Pixel circuitry Active pixels Global shutter pixels Performance Optimization 3 No Security: Public Information Defining the Future of Digital Imaging cmos Image Sensor (CIS) Architecture Charge is not transferred outside the Pixel area Multiple functions integrated with the Sensor array such as amplification, CDS, ADC, readout sequencing and digital processing 4 No Security.

2 Public Information Defining the Future of Digital Imaging CIS Performance Parameters Quantum efficiency (QE) Modulation transfer function (MTF) / spatial resolution Read noise Conversion gain Full well capacity Dark current Lag Shutter efficiency 5 No Security: Public Information Defining the Future of Digital Imaging Photodiodes Advantages High FWC High QE Disadvantages High dark current Low conversion gain No in Pixel charge transfer 6 No Security: Public Information Defining the Future of Digital Imaging Photogates Advantages In Pixel charge transfer Low dark current for buried channel Disadvantages Low blue QE for FSI operation Charge transfer lag (read out time) 7 No Security: Public Information Defining the Future of Digital Imaging Pinned Photodiodes Advantages In Pixel charge transfer Low dark current High QE Disadvantages Lower FWC Charge transfer lag (read out time) 8 No Security.

3 Public Information Defining the Future of Digital Imaging TCAD Device Design Process and device simulation to optimize the Sensor performance (FWC, lag/charger transfer speed) Typical Design parameters that are optimized include implant doses and energies 9 No Security: Public Information Defining the Future of Digital Imaging Light Gathering Micro-lenses Focus light on photodetector Increase effective fill factor Reduce optical crosstalk Anti-reflective coatings = 2 1 2+ 12 1= 2 0 10 No Security: Public Information Defining the Future of Digital Imaging Frontside and Backside Illumination FSI structure works well for larger pixels > 2-3um, but suffers from low QE and high Pixel crosstalk as Pixel size shrinks BSI always has better QE and less optical crosstalk than FSI 11 No Security.

4 Public Information Defining the Future of Digital Imaging Deep Trench Isolation and Buried Color Filters DTI is used to reduce substrate carrier diffusion and therefore increase MTF BCFA is used to reduce optical crosstalk and improve MTF PD PD DTI BCFA CMG 12 No Security: Public Information Defining the Future of Digital Imaging FDTD Optical Simulation Finite difference time domain electro-magnetic equation solver Critical for pixels as their size becomes similar to the wavelength of the illumination Optical confinement methods for continued scaling of cmos Image Sensor pixels C.

5 C. Fesenmaier, Y. Huo, and P. B. Catrysse, Opt. Express 16, 20457 (2008) 13 No Security: Public Information Defining the Future of Digital Imaging * Outline Introduction Photodetectors Pixel circuitry Active pixels Global shutter pixels Performance Optimization 14 No Security: Public Information Defining the Future of Digital Imaging 3T Active Pixel (3T APS) First demonstrated by P. Noble in 1968 High full well capacity High dark current KTC readout noise High speed readout 15 No Security: Public Information Defining the Future of Digital Imaging 3T APS Readout Circuitry 16 No Security: Public Information Defining the Future of Digital Imaging 4T Active Pixel (4T APS) First described with a photo-gate by E.

6 Fossum in 1994 First described with a pinned photodiode by P. Lee in 1997 In Pixel charge transfer enables CDS and removes kTC reset noise Low dark current due to buried channel photodetectors Lower read noise due to separation between photodetector and floating diffusion capacitance Lower fill factor than 3T APS 17 No Security: Public Information Defining the Future of Digital Imaging 4T APS Readout 18 No Security: Public Information Defining the Future of Digital Imaging Active Pixel First demonstrated by M. Mori in 2004 Reduced Pixel size Increased fill factor Slower readout speed (1/2) Higher read noise due to shared floating diffusion 19 No Security: Public Information Defining the Future of Digital Imaging Rolling/Global Shutter Operation Global Shutter Rolling Shutter 20 No Security: Public Information Defining the Future of Digital Imaging Rolling Shutter Artifacts 21 No Security.

7 Public Information Defining the Future of Digital Imaging Voltage Mode Global Shutter Pixel 8T / 2C Pixel [Meynants 15] High shutter efficiency typically ~ 80dB+ KT/C noise limits low light performance due to size of C1 and C2 Large Pixel size / low fill factor 22 No Security: Public Information Defining the Future of Digital Imaging Stacked Voltage Mode Global Shutter Pixel Pixel [Kondo 15] Very high shutter efficiency > 120dB 23 No Security: Public Information Defining the Future of Digital Imaging Charge Mode Global Shutter (Gate Storage) Smaller Pixel size than voltage mode GS [Meynants 15] In Pixel charge transfer allows for CDS and complete kTC noise suppression Buried channel storage gate is needed for low dark current operation Light Shield Storage Gate 24 No Security: Public Information Defining the Future of Digital Imaging * Outline Introduction Photodetectors Pixel circuitry Active pixels Global shutter pixels Performance Optimization 25 No Security.

8 Public Information Defining the Future of Digital Imaging Quantum Efficiency Optimization Frontside / backside illumination EPI thickness Anti reflective coatings Color filters materials Pixel size Micro-lenses 26 No Security: Public Information Defining the Future of Digital Imaging Quantum Efficiency FSI/BSI and EPI Thickness Thicker EPI improves NIR QE but reduces MTF BSI has better QE than FSI 27 No Security: Public Information Defining the Future of Digital Imaging MTF Optimization Frontside / backside illumination EPI thickness Photodetector depletion depth Illumination wavelength Pixel size Buried color filters Deep trench isolation System optics 28 No Security: Public Information Defining the Future of Digital Imaging MTF FSI/BSI and EPI Thickness Thicker EPI reduces MTF 29 No Security.

9 Public Information Defining the Future of Digital Imaging MTF FSI/BSI and Illumination Wavelength MTF is better for short wavelengths for FSI MTF is better for longer wavelengths for BSI MTF of FSI is typically better than BSI 30 No Security: Public Information Defining the Future of Digital Imaging Read Noise Optimization In a well designed CIS the read noise is limited by the transistor connected to the FD When read noise is limited by the transistor connected to the FD then it is proportional to the total input capacitance [Centen 91] 2 = + + 2 ( ) ( )2 Read noise is also limited by the read out bandwidth and the excess or 1/f noise of the input transistor 31 No Security.

10 Public Information Defining the Future of Digital Imaging Read Noise Optimization II Therefore higher conversion gain typically leads to lower read noise, but this limits dynamic range and full well capacity Read noise is not a single parameter, it is a distribution 32 No Security: Public Information Defining the Future of Digital Imaging Full Well Capacity Optimization Surface mode photodetectors have higher capacitance and therefore higher FWC (q=cv) Photodiodes Surface mode photo-gates Larger pixels have higher potential FWC Lower conversion gain enables higher FWC Large FWC is at odds with low read noise and low dark current 33 No Security.