Imaging Resolution

In Search of the Ultimate Chip makers contemplate the 100-megapixel image sensor. Physics says it’s possible, but is it practical or even desirable?

by Brian L. Benamati

he mainstream in image serve. For the astronomy market, veniences of digital image capture. sensors for high-end stu- where scientists work with multi- The advent of digital imaging dio has been million-dollar optics, there will be brought forth image sensor alterna- 6-megapixel CCDs, and one answer. At the other end of the tives in a variety of optical formats: T the market is now adopt- spectrum, where consumers expect 1/5 in. (and smaller) to 2/3 in. for con- ing 16.8-megapixel image sensors as to pay less than $100 for a sumer ; 4/3 in. for so-called the new reference standard. The for the Web, the answer is quite dif- pro-sumer cameras; APS format for semiconductor processes used to ferent. advanced amateur cameras; and 35 manufacture sensors can achieve in- Consider professional photogra- mm for professional cameras. Sensor creasingly smaller pixels and larger phers, who traditionally rely on large- formats of 48 36 mm and “true sensors for even higher-den- 645” (56 42 mm) may be sity chips. The market re- on the horizon. quirements for image quality and the practical considera- Pixel size tions of camera systems, how- Today numerous compa- ever, must be understood to nies offer answer the question: What backs for medium- and large- will be the “ultimate” image format cameras using tradi- sensor of days to come? tional 35-mm-format CCD Marketplace dynamics will image sensors with 6 mega- strongly drive the evolution of pixels, based on 12-µm pix- image sensors. It’s relatively els. The cameras and lenses, easy to conceive of imagers however, are crying out: approaching 100 megapixels “More pixels!” that deliver extraordinary Most recently, the high end Figure 1. A 16-megapixel, full-frame CCD image sensor image quality, but could such of the is today’s reference standard for commercial imagers. devices achieve commercial market began migrating from Physicists can easily conceive of 100-megapixel devices, success? Chip makers face 6 megapixels to a higher-res- but the marketplace has its own conceptions about digital such concerns as process olution 16.8-megapixel (4k imaging cost/performance trade-offs. technology, pixel and chip 4k) image sensor. This de- size, and the number of chips vice uses a 9-µm pixel in a per wafer. Camera makers are con- format (4 5 in.) and so-called “645” 36 36-mm (1:1 ) opti- cerned with sensitivity (effective film medium-format film cameras to cal format. speed), dynamic range, noise, com- achieve exceptional image quality, Over the next several years, image patibility with camera formats and well beyond that of 35-mm-format sensors will transition to 20 mega- legacy lenses, and, of course, price. cameras. Digitizing these larger-for- pixels and even 36 megapixels. By increasing the active sensing mat systems demands larger-format Manufacturers may offer even area and the number of pixels in it, silicon image sensors beyond the tra- higher-resolution solutions, but they higher-resolution sensors improve ditional 35-mm format. Photogra- will appear first in niche markets. image quality, but the level of sensor phers expect these larger image sen- Only time and price will determine performance most suitable for digi- sors to provide quality approaching how broad the market appeal will be tal imaging applications depends on that of traditional large-format pho- for large-format, ultrahigh-resolu- the imaging market segment they tographic film, while offering the con- tion image sensors. Reprinted from the September 2001 issue of PHOTONICS SPECTRA © Laurin Publishing Imaging Resolution

1/ 2/ 3 in. 3 in. 1/ 1/ 4 in. 2 in. 48 36 mm 4/ 1 in. 3 in. APS (3:2) 35 mm 645 mm

Figure 2. Photographers want digital imagers to be compatible with traditional lenses and cameras, so manufacturers try to make the largest possible sensor to fill a camera’s optical format. Smaller sensors change the effective magnification factor of the lens.

Scaling from a 9-µm to a 6.8-µm Common Image Sensor Optical Formats pixel, for instance, would produce Imager Aspect Width Height Diagonal more than 36 megapixels in a 48 Format Ratio H (mm) V (mm) (mm) 36-mm sensor. Development would 1/4 in. 4:3 3.2 2.4 4 not be trivial, but such an image sen- 1/3 in. 4:3 4.8 3.6 6 sor is a realistic proposition. 1/2 in. 4:3 6.4 4.8 8 To take this approach a step or 2/3 in. 4:3 8.8 6.6 11 two further, imagine scaling to a 5- 1 in. 4:3 12.8 9.6 16 µm pixel and increasing optical for- 1 4/3 in. 4:3 17.8 13.4 22.3 mat to 56 42 mm. This would en- APS2,3 3:2 25.1 16.7 30.1 able nearly 100 megapixels, theo- 35 mm 3:2 36.0 24.0 43.3 retically viable for a 645-format 48 36 mm 4:3 48.0 36.0 60.0 camera and lens system. 645 4:3 56.0 41.5 69.7 Tomorrow’s sensor designers and manufacturing processes can gen- 1. KAF-5100CE image sensor erate these ultrahigh-resolution sen- 2. D1: 23.7 (H) 15.6 mm (V), 28.4 mm diagonal sors, but what is really practical? Is 3. Canon D30: 22.7 (H) 15.1 mm (V), 27.3 mm diagonal there a point at which the reduction in pixel size runs into diminishing returns because of the fundamental And chip makers will continue to and slower image capture rates from limitations in sensor sensitivity, push the envelope for pixel perfor- ultrahigh-resolution digital cameras charge capacity or dynamic range? mance as seen in the eyes of the could adversely affect their produc- Generally speaking, an imager’s camera maker and professional pho- tivity. Image sensors beyond 16 ability to gather light and to store tographer. Why should these ad- megapixels can meet the minimum charge at each picture element is a vances cease? frame-rate requirement by using function of a pixel’s effective area. In In the high-volume consumer mar- multiple-output sensors; however, a simple model, a 10-µm pixel (area ket, where the 1/2-in. optical format is this approach creates system-level of 100 µm2) should have twice the the most prevalent, leading-edge res- complexities for the camera maker. “performance” of a 7-µm pixel (area olution now ranges between 3 and 5 of 49 µm2). However, chip makers megapixels, with pixel size decreas- The role of tradition use the best available technologies ing from 4 toward 3 µm. But camera Format compatibility with tradi- to improve a detector’s quantum ef- users have reported that shrinking tional lenses and cameras also plays ficiency (ability to convert incident pixels sometimes compromises image an important role in the professional photons to electrons) and linear quality because of the aforemen- photography segment. charge capacity (ability to correctly tioned fundamental limitations. For digital imaging, it’s best to have store the converted charge) while The system implications for in- the largest-possible sensor to fill the minimizing noise (e.g., dark current creased resolution are also costly. A camera’s optical format, and larger generated in the bulk silicon). higher pixel count means a need for sensors are the logical next step for In consumer electronics, such in- higher-capacity memory and faster the industry, packing more pixels novations are largely responsible for signal processing to achieve compa- into a sensor using an already proven today’s handheld camcorders with rable frame rate, for example. pixel size. “0-lux” performance as compared In the professional market seg- Larger chips, however, present a with older over-the-shoulder cam- ment, trade-offs between higher res- manufacturing challenge. If doubling corders with “4-lux” performance. olution and frame rate are now be- sensor size halves the number of sen- New semiconductor technologies coming apparent. Fashion photog- sors on a wafer, the cost to manu- have improved the detector’s perfor- raphers, for example, develop a facture the chip essentially doubles. mance per nano-acre substantially. rhythm while working with models, Managing yield for such devices to Imaging Resolution

world may be flooded with Image Sensor Resolution vs. Format Diagonal lenses designed for digital 1/ in. imaging. But the practical re- 2 4 1 2 /3 in. 48 36 mm 1 /3 in. /3 in. ality is that imager manufac- /4 in. 1 in. APS 35 mm 645 mm turers must serve traditional 40 cameras and lenses. Given 35 that the resolving power of tra- Pixel Size 5 µm Studio ditional professional lenses is 30 6.8 µm about 5 µm, moving to smaller Professional pixels — such as those in con- 25 µ Advanced 9 m sumer sensors — will not provide an effective resolution gain and would Amateur 11 µm 20 likely pose a problem. Pro-Sumer 13 µm 15 The ultimate sensor 10 Consumer As digital imaging science advances and image sensor design and pro- 5 cessing technologies improve, the

Imager Resolution (megapixels) professional photography market will 0 benefit from larger-format chips with 0 10 20 30 40 50 60 70 80 smaller pixel sizes. Chip sizes of 48 Format Diagonal (mm) 36 mm and 52 42 mm would fulfill the optical requirements of pro- Figure 3. Improving resolution for a given sensor format/size means shrinking fessional photographers. And as pixel pixels, which can affect the sensor’s ability to gather light and store charge. architectures migrate from 12 to 9 µm, and then toward 7 and even 5 achieve a practical price point is a Smaller image sensors may seem µm, the chips will provide extraor- major challenge. more economically appealing, but dinary resolution. Other challenges of creating larger they cause the camera maker signi- These technological advances will sensor chips include achieving flat- ficant optical problems. In particular, bring the ultimate in performance to ness and photoresponse uniformity a sensor that’s much smaller than digital cameras. The true limitation across large dies. For the camera the optical field presented to the focal will not be the chip maker’s ability to maker, it can be a challenge to focus plane of the lens changes the effec- deliver image sensors, but rather the such a large chip within a camera tive magnification factor of the lens practical trade-off between image back on a traditional camera. so that what you see through the quality and price that the photo- Larger sensors also pose a prob- viewfinder is not what you capture. graphic market is willing to bear. lem for conventional semiconductor Professional photographers will not Whatever the marketplace decides, manufacturing techniques. Manufac- accept large photographic magnifica- chip makers will strive to surpass turers who use steppers for lithog- tion errors. A 35-mm-format (36 24 the wildest dreams of camera mak- raphy, for example, are limited by mm) sensor has a mag- ers and photographers, who are all the field size of the stepper, and some nification error of 1.62 for a true on the same quest: the search for have been forced to use “stitching” 645 (56 42 mm) field. A 48 36- the ultimate image sensor. G to make larger chips. This forces the mm sensor would have a 1.16 fac- camera maker to deal with stitch tor. A true 645 sensor (56 42 mm) Meet the author boundaries and to minimize bound- would have “perfect” (1.00) magni- Brian L. Benamati is a product mar- ary errors in the final image. Newly fication in such a camera, but pe- keting engineer in the Image Sensor introduced semiconductor equipment ripheral fading caused by lens roll-off Solutions Div. of Eastman Kodak Co. in should be able to avoid these issues near the edges of the field would be Rochester, N.Y. He received BS and MS in manufacturing the next genera- of concern. degrees in electrical engineering from the tion of large-format image sensors. In the future, the photographic Rochester Institute of Technology.