Sensor nonlinearity

The conversion of light into photocharge is a highly linear process.

In silicon, this has been verified for a large dynamic range of at least 10orders of magnitude [14]. Unfortunately, much of this linearity is lost in the photocharge detection principle that is mainly used in image sensors. Photocharge is stored as the state of discharge of a precharged capacitance, either an MOS capacitance or a photodiode. As the width of the space-charge region depends on the discharge level, the spectral sensitivity and the photometric linearity are a function of the amount of photocharge already stored.

The same problem is encountered in the electronic charge detection circuits that are implemented as source followers after a floating diffu-sion (see Fig.5.15). The capacitance of the floating diffusion depends on the voltage on it and therefore on the charge state. This causes nonlinearities in charge sensing.

The degree of the nonlinearity depends very much on the charge de-tection (or voltage) range that is used. For differential measurements of over a few hundred mV in the middle region of the analog sensor output, nonlinearities can be below 0.1 % [45]. Over the full sensing

Vbias

C reset

Vout

Figure 5.22:Schematic diagram of a charge detection circuit, providing a high photodetection linearity by keeping the photodiode voltage constant. If the feed-back capacitance is replaced by a resistor, a so-called transimpedance amplifier results, converting photocurrent in a proportional voltage with very high linear-ity.

range, nonlinearities may be as large as a few percent. If the mea-surement should be highly linear, a proper electronic charge detector circuit must be used in which the voltage at the input is kept constant.

Such a charge detector circuit, illustrated in Fig.5.22, requires a cer-tain amount of silicon floorspace. With state-of-the-art semiconductor technology, pixels become so large that only 1-D arrays have been real-ized with this technique [46]; in image sensors it is not yet realistic to implement such charge detectors in each pixel. For this reason, image sensing applications for optical metrology in which sub-percent lin-earity is demanded have to resort to accurate calibration and off-chip digital correction techniques [5].

5.9 Conclusions

It was only about a decade ago that a few researchers started to exploit one of the most exciting capabilities offered by modern silicon-based semiconductor technology, the monolithic integration of photosensi-tive, analog and digital circuits. Some of the results of these efforts are described in this work, representing just a small fraction of the many applications already demonstrated. They all support the main asser-tion of this chapter, that today’s image sensors are no longer restricted to the acquisition of optical scenes. Image sensors can be supplied with custom integrated functionality, making them key components, application-specific for many types of optical measurement problems.

It was argued that it is not always optimal to add the desired custom functionality in the form of highly-complex smart pixels, because an in-crease in functionality is often coupled with a larger fraction of a pixel’s area being used for electronic circuit, at the cost of reduced light sen-sitivity. For this reason, each new optical measurement problem has

5.10 References 149 to be inspected carefully, taking into account technical and economical issues. For optimum system solutions, not only smart pixels have to be considered. Functionality could also be provided by separate on-chip or off-chip circuits, perhaps by using commercially available electronic components.

Machine vision system architects can no longer ignore the freedom and functionality offered by smart image sensors, while being well aware of the shortcomings of semiconductor photosensing. It may be true that the seeing chips continue to be elusive for quite some time.

The smart photosensor toolbox for custom imagers is a reality today, and a multitude of applications in optical metrology, machine vision, and electronic photography can profit from the exciting developments in this area. “Active vision,” “integrated machine vision,” “electronic eyes,” and “artificial retinae” are quickly becoming more than concepts:

the technology for their realization is finally here now!

5.10 References

[1] Gonzalez, R. and Wintz, P., (1987). Digital Image Processing, 2nd edition.

Reading, MA: Addison-Wesley.

[2] Beck, R. (ed.), (1995).Proc. AAAS Seminar on Fundamental Issues of Imag-ing Science, Atlanta (GA), February 16-17, 1995.

[3] Beyer, H., (1992). Geometric and radiometric analysis for a CCD-camera based photogrammetric close-range system. PhD thesis No. ETH-9701, Federal Institute of Technology, Zurich, Switzerland.

[4] Chamberlain, S. and Lee, J., (1984). A novel wide dynamic range silicon photodetector and linear imaging array.IEEE Jour. Solid State Circ.,SC-19:

175–182.

[5] Lenz, R., (1996). Ein Verfahren zur Schätzung der Parameter ge-ometrischer Bildtransformationen. Dissertation, Technical University of Munich, Munich, Germany.

[6] Schenker, P. (ed.), (1990). Conference on Active Vision, Vol. 1198 ofProc.

SPIE.

[7] Seitz, P., (1995). Smart image sensors: An emerging key technology for ad-vanced optical measurement and microsystems. InProc. SPIE, Vol. 2783, pp. 244–255.

[8] Saleh, B. and Teich, M., (1991). Fundamentals of Photonics. New York:

John Wiley and Sons, Inc.

[9] Wong, H., (1996). Technology and device scaling considerations for CMOS imagers. Vol. 43, pp. 2131–2142. ISPRS.

[10] Sze, S., (1985).Semiconductor Devices. New York: John Wiley and Sons.

[11] Spirig, T., (1997). Smart CCD/CMOS based image sensors with pro-grammable, real-time temporal and spatial convolution capabilities for applications in machine vision and optical metrology. PhD thesis No. ETH-11993, Federal Institute of Technology, Zurich, Switzerland.

[12] Heath, R., (1972). Application of high-resolution solid-state detectors for X-ray spectrometry—a review.Advan. X-Ray Anal.,15:1–35.

[13] Bertin, E., (1975).Principles and Practice of X-Ray Spectrometric Analysis.

New York: Plenum Press.

[14] Budde, W., (1979). Multidecade linearity measurements on Si photodi-odes.Applied Optics,18:1555–1558.

[15] Theuwissen, A., (1995).Solid-State Imaging with Charge-Coupled Devices.

Dordrecht, The Netherlands: Kluwer Academic Publishers.

[16] Vietze, O. and Seitz, P., (1996). Image sensing with programmable offset pixels for increased dynamic range of more than 150dB. InConference on Solid State Sensor Arrays and CCD Cameras, Jan. 28–Feb. 2, 1996, San Jose, CA, Vol. 2654A, pp. 93–98.

[17] Vietze, O., (1997). Active pixel image sensors with application specific performance based on standard silicon CMOS processes. PhD thesis No.

ETH-12038, Federal Institute of Technology, Zurich, Switzerland.

[18] Webb, P., McIntyre, R., and Conradi, J., (1974). Properties of Avalanche Photodiodes.RCA Review,35:234–277.

[19] Seitz, P., (1997). Image sensing with maximum sensitivity using industrial CMOS technology. InConference on Micro-Optical Technologies for Mea-surement Sensors and Microsystems II, June 16–June 20, 1997, Munich, Germany, Vol. 3099, pp. 22–33.

[20] Zappa, F., Lacatia, A., Cova, S., and Lovati, P., (1996). Solid-state single-photon detectors.Optical Engineering,35:938–945.

[21] Mathewson, A., (1995). Integrated avalanche photo diode arrays. Ph.D.

thesis, National Microelectronics Research Centre, University College, Cork, Ireland.

[22] Mahowald, M., (1991). Silicon retina with adaptive photodetectors. In Conference on Visual Information Processing: From Neurons to Chips Jan.

4, 1991, Orlando, FL, Vol. 1473, pp. 52–58.

[23] Graf, H., Höfflinger, B., Seger, Z., and Siggelkow, A., (1995). Elektronisch Sehen. Elektronik,3:3–7.

[24] Sankaranarayanan, L., Hoekstra, W., Heldens, L., and Kokshoorn, A., (1991). 1 GHz CCD transient detector. InInternational Electron Devices Meeting 1991, Vol. 37, pp. 179–182.

[25] Colbeth, R. and LaRue, R., (1993). A CCD frequency prescaler for broad-band applications. IEEE J. Solid-State Circ.,28:922–930.

[26] Carnes, J. and Kosonocky, W., (1972). Noise sources in charge-coupled devices. RCA Review,33:327–343.

[27] Allen, P. and Holberg, D., (1987).CMOS Analog Circuit Design. Fort Worth:

Saunders College Publishing.

[28] Hopkinson, G. and Lumb, H., (1982). Noise reduction techniques for CCD image sensors.J. Phys. E: Sci. Instrum,15:1214–1222.

[29] Knop, K. and Seitz, P., (1996). Image Sensors. InSensors Update, W. G.

Baltes, H. and J. Hesse, eds., pp. 85–103. Weinheim, Germany: VCH-Verlagsgesellschaft.

5.10 References 151 [30] Chandler, C., Bredthauer, R., Janesick, J., Westphal, J., and Gunn, J., (1990).

Sub-electron noise charge coupled devices. In Conference on Charge-Coupled Devices and Solid State Optical Sensors, Feb. 12–Feb. 14, 1990, Santa Clara, CA, Vol. 1242, pp. 238–251.

[31] Janesick, J., Elliott, T., Dingizian, A., Bredthauer, R., Chandler, C., West-phal, J., and Gunn, J., (1990). New advancements in charge-coupled device technology. Sub-electron noise and 4096 X 4096 pixel CCDs. In Confer-ence on Charge-Coupled Devices and Solid State Optical Sensors, Feb. 12–

Feb. 14, 1990, Santa Clara, CA, Vol. 1242, pp. 223–237.

[32] Matsunaga, Y., Yamashita, H., and Ohsawa, S., (1991). A highly sensitive on-chip charge detector for CCD area image sensor. IEEE J. Solid State Circ.,26:652–656.

[33] Fossum, E., (1993). Active pixel sensors (APS)—are CCDs dinosaurs? In Conference on Charge-Coupled Devices and Solid-State Optical Sensors III, Jan. 31–Feb. 2, 1993, San Jose, CA, Vol. 1900, pp. 2–14.

[34] Mendis, S., Kemeny, S., Gee, R., Pain, B., Staller, C., Kim, Q., and Fossum, E., (1997). CMOS active pixel image sensors for highly integrated imaging systems.IEEE J. Solid-State Circ.,32:187–197.

[35] DeMarco, P., Pokorny, J., and Smith, V. C., (1992). Full-spectrum cone sensitivity functions for X-chromosome-linked anomalous trichromats.

J. Optical Society,A9:1465–1476.

[36] Wendland, B., (1988).Fernsehtechnik I: Grundlagen. Heidelberg: Hüthig.

[37] Pratt, W., (1991).Digital image processing. New York: Wiley.

[38] Committee on Colorimetry, Optical Society of America, (1953). The Sci-ence of Color. Washington, D. C.: Optical Society of America.

[39] Inglis, A. F., (1993).Video engineering. New York: McGraw-Hill.

[40] Pritchard, D. H., (1984). U.S. color television fundamentals — a review.

RCA Engineer,29:15–26.

[41] Hunt, R. W. G., (1991).Measuring Colour,2nd edition. Ellis Horwood.

[42] Bayer, B. E., (1976). Color imaging array, U.S. patent No. 3,971,065.

[43] Knop, K., (1985). Two-dimensional color encoding patterns for use in single chip cameras.Proc. SPIE,594:283–286.

[44] Aschwanden, F., Gale, M. T., Kieffer, P., and Knop, K., (1985). Single-chip color camera using a frame-transfer CCD. IEEE Trans. Electron. Devices, ED-32:1396–1401.

[45] Flores, J., (1992). An analytical depletion-mode MOSFET model for analy-sis of CCD output characteristics. InConference on High-Resolution Sen-sors and Hybrid Systems, Feb. 9–Feb. 14, 1992, San Jose, CA, Vol. 1656, pp. 466–475.

[46] Raynor, J. and Seitz, P., (1997). A linear array of photodetectors with wide dynamic range and near photon quantum noise limit. Sensors and Actuators A,61:327–330.

6 Geometric Calibration of Digital