Manufacturers must consider the trade-offs in the different levels of sensor integration to offer the right CMOS camera for the application. While the CCD offers the best sensor performance for digital imaging, the CMOS image sensor is superior in terms of integrating functionality on-chip. Even the much-touted camera-on-a-chip is not beyond the capabilities of CMOS technology, promising sensors that incorporate drivers, power supplies, timing generators and even correlated double-sampling units, analog-to-digital converters and digital signal processors on-chip. But just because certain circuitry can be integrated onto a sensor chip, it does not follow that is always advantageous to do so. Therefore, the optimal degree of integration for today’s — as well as tomorrow’s — applications will continue to differ according to the desired application. CMOS camera manufacturers must consider the target application when deciding how much integration is necessary. Although additional circuitry can be integrated onto the sensor, it is not necessarily advantageous to do so. Photos courtesy of Eastman Kodak Co. The so-called single-chip camera offers the physically smallest digital camera design and the lowest possible cost and parts count, but there are limitations to its performance. Moreover, its economic and real estate advantages are not as great as they at first may appear. For both analog-to-digital conversion and digital signal processing, the expertise of the companies that specialize in these areas typically exceeds that of sensor manufacturers, and the discrete integrated circuits typically have tighter linewidth geometries (0.13 to 0.18 μm) than the sensor makers’ integrated imagers (0.35 to 0.5 μm). It is possible, of course, to shrink the geometries of image sensors, but this tends to diminish their performance and image quality, particularly with respect to dark current (higher noise) and charge capacity. Quality consumer cameras will employ a two-chip design, combining an integrated CMOS sensor and a discrete digital signal processor. An integrated analog-to-digital converter or digital signal processor always requires less space than a discrete one but is not necessarily less expensive. The dense integration of discretes will keep the die size small on the silicon wafer, yielding major cost savings as well as minimizing the required space for the packaged component. On the other hand, adding functions to sensor chips increases their die size and compromises their real estate and cost advantage. For applications that demand the highest performance and image quality, it is advantageous to implement certain functions with discrete, state-of-the-art integrated circuits. The best design for quality consumer cameras, for example, will be a two-chip device combining an integrated CMOS sensor that contains all the functionality on-chip for acquiring an image and a discrete digital signal processor that offers color correction, white balancing and such signal-processing functions as noise filtering and, perhaps, defect correction. “Prosumer” and professional applications will require a three-chip design, comprising an integrated CMOS sensor, a discrete digital signal processor and, to attain the best noise performance, a discrete analog-to-digital converter. Single-chip CMOS cameras will find applications in cellular phones and in a variety of other handheld consumer and communications electronic devices into which digital imaging will be integrated, serving where size, and not quality, is the primary concern. For applications where image quality is the thing, the future belongs to multichip CMOS designs and — at the highest end of the market — multichip designs based on CCDs. Meet the author Helen Titus is marketing manager at Eastman Kodak Co., Image Sensor Solutions, in Rochester, N.Y. She holds a master’s and a bachelor’s degree in electrical engineering from?Rochester Institute of Technology and from Gannon University in Erie, Pa., respectively