Prof. dr. ir. Albert J.P. THEUWISSEN Philips Semiconductors Image Sensors, Eindhoven (NL) Technical University, Delft (NL)

Transcription

1 CCD or CMOS image sensors for consumer digital still photography? Prof. dr. ir. Albert J.P. THEUWISSEN Philips Semiconductors Image Sensors, Eindhoven (NL) Technical University, Delft (NL) Introduction Outline Principle of CCD and CMOS Imager Overview CCD : resolution, signal-to-noise ratio, angular response, dark current, dynamic range, linearity, pixel uniformity, architecture Summary and Conclusions 1

2 Introduction CMOS is challenging CCD Digital still is a continuously growing imaging market Introduction CMOS is challenging CCD Digital still is a continuously growing imaging market Today : almost exclusively CCD in DSC Tomorrow : CCD or CMOS? 2

3 Introduction Outline Principle of CCD and CMOS Imager Overview CCD : resolution, signal-to-noise ratio, angular response, dark current, dynamic range, linearity, pixel uniformity, architecture Summary and Conclusions Φ CCD principle (1) 0 V 10 V 0 V 0 V p-si 0 V 10 V 10 V 0 V Φ p-si 0 V 0 V 10 V 0 V Φ p-si 3

4 CCD principle (2) photosensitive CCD array out horizontal CCD output register CMOS principle (1) vertical scan circuit photodiode array + MOS switches A/D horizontal scan circuit 4

5 CMOS principle (2) Photogate APS Photodiode APS RST RST PG TX RS RS p-si n + Col bus p-si n + Col bus Introduction Outline Principle of CCD and CMOS Imager Overview CCD : resolution, signal-to-noise ratio, angular response, dark current, dynamic range, linearity, pixel uniformity, architecture Summary and Conclusions 5

6 Image Sensor Aspects (1) IMAGER CAMERA PARAMETER SPECIFICATION resolution signal-to-noise ratio angular response dark current sharpness ISO speed min. F-stop max. exp. time Image Sensor Aspects (2) IMAGER CAMERA PARAMETER SPECIFICATION dynamic range linearity pixel uniformity architecture latitude colour fidelity granularity features 6

7 Introduction Outline Principle of CCD and CMOS Imager Overview CCD : resolution, signal-to-noise ratio, angular response, dark current, dynamic range, linearity, pixel uniformity, architecture Summary and Conclusions Resolution Requirements Print Resolution [pix/inch] "x5" 4"x6" 5"x7" 200 8"x10" quality level Image Size [Mpixels] 7

8 Resolution Requirements Print Resolution [pix/inch] "x5" 4"x6" 5"x7" 200 8"x10" quality level disc 110 APS 35 mm Image Size [Mpixels] Resolution Requirements Print Resolution [pix/inch] prof. CCD cons. CCD cons. CMOS prof. CMOS quality level disc 110 APS 35 mm Image Size [Mpixels] 3.5"x5" 4"x6" 5"x7" 8"x10" 8

9 Resolution [Mpixels] Trend in Resolution 10 2/3" (8.8mmx6.6mm) 8 1/2" (6.4mmx4.8mm) 1/3" (4.4mmx3.3mm) 6 decrease chip size Pixel Size [um] Resolution [Mpixels] Trend in Resolution 10 lens limitation 2/3" (8.8mmx6.6mm) 8 1/2" (6.4mmx4.8mm) 1/3" (4.4mmx3.3mm) 6 4 decrease chip size 2 CCD CMOS Pixel Size [um] 9

10 Number of Electrons Signal-to-Noise ratio (1) 100k 10k 1k signal saturation (100k) 100 photon shot total (40) dark read noise (20) shot (35) Light Intensity (µw/cm 2 ) dynamic range (68 db) Signal-to-Noise ratio (2) Photogate APS Photodiode APS RST RST PG TX RS RS n + Col bus n + Col bus p-si p-si FREE of reset noise LOW light sensitivity NOT FREE of reset noise HIGH light sensitivity 10

11 Signal-to-Noise ratio (3) ISO = x 10 H x ISO x = S/N=x H x = exposure to get S/N=x ISO 40 A QE ISO 10 A QE n r A = pixel area QE = quantum efficiency n r = read noise Signal-to-Noise ratio (4) QE in green [%] Pixel Size [um] 11

12 Signal-to-Noise ratio (4) QE in green [%] CCD CCD 20 CMOS Pixel Size [um] Improvement QE (1) microlens pixels 12

13 Improvement QE (2) lens iris microlens pixels Angular Response (1) Incident Angle [deg.] F-number 13

14 Angular Response (2) Rel. Imager Response Incident Angle [deg.] Angular Response (3) Incident Angle [deg.] 25 micro-lens limit F-number 14

15 Dark Current (1) prof. CCD pa/cm RT cons. CCD pa/cm RT standard CMOS pa/cm RT Dark current doubles every 6 8 o C. Example 60 o C : 32 times -100 o C : 32,000 times lower! Dark Current (2) Compensation for dark current is possible! Compensation for dark-current nonuniformities is possible! Compensation for dark-current shot-noise is NOT possible! 15

16 DR = Dynamic Range (1) N sat n 2 r N + n dark 2 dark DR = dynamic range N sat = saturation signal [e - ] N dark = dark signal [e - ] n r = read noise [e - ] n dark = dark shot noise [e - ] Charge Handling [ke/um2] Dynamic Range (2) 3 N sat = 30 ke Pixel Size [um] 16

17 Charge Handling [ke/um2] Dynamic Range (2) 3 N sat = 30 ke CCD 1 CMOS Pixel Size [um] RAW data interpolation Linearity (1) white balance R G = x B colour matrixing R G B γ = 1.8 gamma curve R = R x 1.40 G = G x 1.00 B = B x 1.46 RGB data 17

18 Linearity (2) Due to sampling in colour space : Interpolations, Filters do not match perfectly : Colour corrections. Linearity (2) Due to sampling in colour space : Interpolations, Filters do not match perfectly : Colour corrections. Linearity CCD : 99 % (for 70 % of N sat ), Linearity CMOS : 97 % (for 85 % of N sat ). 18

19 - p.r.n.u. Pixel Random Non- Uniformity PRNU CCD : PRNU CMOS : < %, < %, (column + pixel FPN) - p.r.n.u. Pixel Random Non- Uniformity PRNU CCD : PRNU CMOS : < %, < %, (column + pixel FPN) Can be corrected by means of LUT, To be non-visible : PRNU and FPN < photon shot noise (0.5 % for 40 ke - ). 19

20 - p.r.n.u. - architecture Architecture CCD : parallel integration/reset CMOS : rolling integration/reset - p.r.n.u. - architecture Architecture CCD : parallel integration/reset CMOS : rolling integration/reset Can be solved by 1 T and 1 C extra in every pixel extra Costs sensitivity, charge capacity, noise,... 20

21 - p.r.n.u. - architecture Introduction Outline Principle of CCD and CMOS Imager Overview CCD : resolution, signal-to-noise ratio, angular response, dark current, dynamic range, linearity, pixel uniformity, architecture Summary and Conclusions - p.r.n.u. - architecture - summary Summary (1) Resolution : pixel size of CCD smallest Noise : CMOS pixels suffer from reset noise Quantum efficiency : CMOS and CCD can be similar Angular response : limits set by micro-lenses Dark current : CCD outstanding 21

22 - p.r.n.u. - architecture - summary Summary (2) Saturation level can be similar Dynamic range of CCD is higher Linearity of CCD is better Pixel uniformity of CCD is better Device architecture of CCD gives more flexibility - p.r.n.u. - architecture - summary - Conclusions (1) CCD or CMOS image sensor CCD? YES! for consumer digital still photography? CMOS? YES, provided that noise and dark current problems can be solved!!! 22

23 - p.r.n.u. - architecture - summary - Conclusions (2) Main issue : S/N performance Then benefit from : low power of CMOS, low driving voltages of CMOS, on-chip functionality, selective read-out mechanism, cost advantage. - p.r.n.u. - architecture - summary - - remark Important Remark This presentation was about digital still photography. For video applications the situation changes completely!!! 23

24 - p.r.n.u. - architecture - summary - - remark - references References R. Baer : IEEE workshop on CCD & AIS, Karuizawa, 1999, J. Bosiers et.al. : IEDM, San Francisco, 1998, M. Kriss : ICPS, Antwerp, 1998, A. Theuwissen : Solid-state imaging with Charge- Coupled Devices, 1995 E. Fossum : Camera-on-a-chip, IEEE-ED, Oct.,