IISW 2009 Backside Illumination Symposium

Transcription

1 IISW 2009 Backside Illumination Symposium The Mass Production of BSI CMOS Imager Sensors Dr. Howard Rhodes Omnivision Technologies, Inc. 1 Acknowlegement D. Tai, Y. Qian, D. Mao, V. Venezia, Wei Zheng, Z. Xiong, C.Y. Liu, K.C. Ku, S. Manabe, A. Shah, S. Sasidhar, P. Cizdziel, Z. Lin, A. Ercan, M. Bikumandla, R. Yang, P. Matagne, C. Yang, H. Yang, T.J. Dai, J. Li, S.G. Wuu 1,D.N. Yaung 1, C.C. Wang 1, J.C. Liu1, C.S. Tsai 1, Y.L.Tu 1, T.H. Hsu 1 Omnivision Technologies, Inc. USA 1 Taiwan Semiconductor Manufacturing Company, Taiwan, R.O.C. 2 1

2 Differences between FSI and BSI FSI ML BSI CF Metal layers Metal layers PD Thinner Si Thickness PD P-epi Cross talk P+ P-epi DPW P+ substrate DPW Dark current CF ML ARC Light Reflection 3 Advantages of BSI over FSI Tall metal stack interfering with light penetration is eliminated Potential to achieve very high peak color QE s (> 70 %) High QE s enable improved sensitivity and low light performance Enables any number of metal layers on frontside with no performance penalty Excellent CRA and angular response Low Stack Height enables lower F/# Lenses for increased sensitivity Excellent Image Performance across the entire image plane 100% Fill Factor Ultra thin epi provides improved electrical crosstalk Enables higher performance for pixels 1.75 µm and smaller 4 2

3 Early History: Sept 2006 BSI B&W QE Quantum Efficiency improvement of backside illumination through steady process development 5 Early BSI History: Nov µm BSI Image in Color Early Historical Image with 30 % B&W QE 75 % B&W QE achieved by Dec 2006 but not on this image 6 3

4 Early BSI History: May µm BSI in Color 7 Early BSI History: May µm BSI in Color 8 4

5 Early BSI History: May µm BSI in Color 9 Early History:1.75 µm Pixel Angular Response Comparison 120% 100% % of Center Maximum 80% 60% 40% BSI FSI 20% 0% Angle (deg) Much improved angular response for BSI compared to FSI May

6 Challenges of BSI To marry the FS with the BS processing so that the generated electrons are effectively delivered to the target photodiode with high QE and low crosstalk To develop a low stress, void-free BS bonding of the device wafer to the handle wafer To develop a BS thinning process with tight thickness control To completely passivate the BS surface to achieve low dark current and low white pixel defect density To develop the BS patterning and BS etching to enable a BS metal light shield To develop backside patterning with accurate alignment CFA and microlens to enable a color CIS imager To solve any new reliability issues To do this with a low cost process at high yield 11 BSI Process Flow 12 6

7 1.4µm Pixel Cross-Section 13 3D Simulation Strategy Mask layout 3D device including ARC & ML 3D EM simulation Poynting vector CFA, µlens, focus 2D & 3D Process Simulation Optical Generation Doping Profiles 3D device including Doping 3D Device simulation Photocurrents, crosstalk, QE, blooming Mask layout No Security: Company Public Confidential Information 14 7

8 3D Optical Simulation 3x3 Pixel Array Com pany Confi 15 3D Electrical Device Simulation Com pany Confi 16 8

9 1.4 µm Pixel Blooming FWC Illuminated pixel dark pixel Xtalk Blooming E µm Pixel: QE vs Wavelength 9.00E E E E E E+01 Simul-QE (B&W) Simul-QE(B) Simul-QE(G) Simul-QE(R) Data-QE(B) Data-QE(R) Data-QE(G) 3.00E E E E

10 Quantum Efficiency R, Gb, Gr, and B Channels vs Wavelength for 1.4µm Pixel 19 Color Accuracy 100lux 1.4µm Pixel Image Using 3200K light Source deltae : Total Color Difference deltac : Chroma Error deltah : Hue Error Ideal Real CCM using D65 as Target 20 10

11 1.4 µm Pixel Raw Black Level Black level is very stable across the Image (Pedestal is set to 4 DN) 21 Absolute Conversion Gain Fe55 Kα X-ray Histogram 1.4µm BSI Pixel DN = 1620e FWHM = 26e Kα Kβ

12 White Color Ratios for 1.4µm Pixel 23 Luma S/N after CCM and AWB vs Light Level (Lux) 24 12

13 Measured Gb-Gr Difference for 1.4µm Pixel 25 Lag(e) vs Signal Level(e) for 1.4µm pixel 26 13

14 Raw Dark Image 1.4µm Pixel Product Analog Gain = 8X Digital Gain = 32X 27 Dark Current vs Temperature 28 14

15 Dark Current Histogram for 1.4µm pixel at 50 C 29 8 Megapixel at 100 Lux (15 fps) 1.4µm Pixel 30 15

16 1.4µm Pixel at 300 Lux, 15 fps A Light D65 Light µm Pixel at 100 Lux, 15 fps A Light D65 Light 32 16

17 1.4µm Pixel at 30 Lux, 15 fps A Light D65 Light µm Pixel at 10 Lux, 15 fps A Light D65 Light 34 17

18 Visual Blooming Performance: 1.4µm Pixel 35 Quantitative Blooming Performance: 1.4µm Pixel Image1 Image2 Image3 Image2 - Image3 1X FWC TX Normal Operation 30X FWC TX Normal Operation 30X FWC TX leaking for Blooming control Electrical Blooming is X FWC Condition 36 18

19 1.4µm and 1.75µm Pixel Performance Parameter 1.4 µm 1.75 µm Parameter 1.4 µm 1.75 µm FWC 4500 e 6500e B/G 31.80% 24.0% Peak QE - R 43.8% 53.0% G/R 35.20% 28.9% Peak QE - Gb 53.5% 60.1% B/R 7.30% 5.4% Peak QE - Gr 53.6% 60.2% Peak QE - B 51.6% 60.4% Sensitivity (530nm) 671 mv/lux-sec 1500 mv/lux-sec Read Noise (pixel/periph) 1.9/1.3 e 2.1e/1.3e R/B 5.70% 4.8% PRNU 0.75% 0.8% G/B 14.40% 11.4% R/G 11.20% 9.2% Lux for 10:1 SNR 110 Lux 60 Lux Dark Current (50 C) 27 e/sec 22 e-/s 37 Quantum Efficiency R, Gb, Gr, and B Channels vs Wavelength for 1.75µm Pixel 38 19

20 Color Accuracy 100lux 1.75µm Pixel Image Using 3200K light Source deltae : Total Color Difference deltac : Chroma Error deltah : Hue Error CCM using D65 as Target Ideal Real 39 Raw Black Level for 1.75 µm Pixel Black level is very stable across the Image Pedestal Value is set to 3.5 bits in 8 bit domain 40 20

21 1.75µm BSI Imager Raw, Interpolated Images at 4 Different Lighting Conditions A Light Day Light Cool White TL84 Because of the low color cross talk, OmniBSI pixel will show exceptional color reproduction regardless of the light source. 41 Dark Current Histogram for 1.75µm pixel at 50 C 42 21

22 Visual Blooming Performance: 1.75 µm Pixel 43 Quantitative Blooming Performance: 1.75 µm Pixel Image1 Image2 Image3 Image2 - Image3 1X FWC TX Normal Operation 30X FWC TX Normal Operation 30X FWC TX leaking for Blooming control Electrical Blooming is ~ 30X FWC Condition 44 22

23 OmniBSI TM Reliability Test Summary 45 OmniBSI TM FIT RATE and MTBF Calculations 46 23

24 Conclusions BSI Performance advantages achieved at 1.4 µm and 1.75 µm pixel nodes P/P+ Substrates enable Low Cost BSI Manufacturing Manufacturing Issues Solved No Reliability Issues found with OmniBSI TM Architecture Three 1.4 µm pixel products and three 1.75 µm pixel products all achieving the same predictable performance demonstrates a BSI process that is under control and in mass production BSI is an enabling technology for pixels 1.4 µm and smaller The next shrink of imager design rules will enable a 1.1 µm pixel with reasonable performance Sub-Micron Pixel (SMP) Technology is the next major challenge 47 References 1 Peter Noble, IEEE Trans. El Dev. 15, p , S. Chamberlain, IEEE J Solid-State Circuits SC-4 (6) p , P. Weimer et al., IEEE Spectrum 6 (3): p , P. Denyer et al, VSLI, p , E.R. Fossum Proc. SPIE Vol 1900, p. 2-14, H. Rhodes et al., Workshop on Microelectronics and Electron Devices, p. 7-18, C-R. Moon et al, IEDM Tech. Dig, p , J. Prima et al., Int. Image Sensor Workshop, p. 5-8, T. Joy et al., IEDM Tech. Dig., p , J. Ahn et al., IEDM Tech. Dig. p , Y.Wu, P. Cizdziel, H. Rhodes, SPIE Electronic Imaging Conference, J. Alakarhu, Int. Image Sensor Workshop, p. 1-4, X. Wang et al, Int. Image Sensor Workshop, p , J. Ahn et al, IEDM Tech. Dig., p , M. Ihama et al, Proc. SPIE, p. 6656,

25 Thank You 49 25