Foveon to challenge with 16 megapixels
Just two weeks ago Kodak announced a range of new / improved CCD's. Today the NY Times (online) is carrying a story of a new prototype camera from Foveon which also sports a 16 megapixel imager, this time a CMOS device at a considerably lower price (according to them) than the Kodak device. They say "...able to capture digital images with a resolution of 4,096 by 4,096 picture elements - or pixels - per square inch. That, by some measures, is about twice the resolution of 35-millimeter film."
Low-Price, Highly Ambitious Digital Chip
Suddenly the future of digital photography seems to be becoming much clearer.
Only two weeks ago, Eastman Kodak announced a chip able to capture digital images with a resolution of 4,096 by 4,096 picture elements or pixels per square inch. That, by some measures, is about twice the resolution of 35-millimeter film.
Today, a company founded by one of Silicon Valley's pioneer chip designers, will announce an image-sensing chip capable of the same resolution as the Kodak chip, but made using a technique that could be much less expensive.
Executives of the company, Foveon, said they had given a prototype camera based on their chip to a photographer in Los Angeles, Greg Gorman, who had used it to make a portrait of a cowboy. In that image, no pixels, or dots, were visible to the eye, even with the photograph blown up to a size of 8 feet by 4 feet.
Already, digital cameras being sold on the consumer market for less than $1,000 are rivaling 35-millimeter film cameras. Digital images of the clarity achieved with Foveon chip could begin to challenge even the much more expensive 4-by-5 film cameras made by companies like Hasselblad that are used by professional photographers for portraiture, advertising and fashion.
"We're headed to flat-out replace the film camera," said Carver Mead, the founder of Foveon, which is based in Santa Clara, Calif. Mr. Mead, a pioneer of the chip industry, became a Silicon Valley legend in the 1970's by helping develop techniques that for the first time enabled chip engineers to create circuits containing tens of thousands of transistors.
Industry analysts say that the new technologies could affect much more than still cameras. High-resolution images, if produced in quantities that made the new generation of image-sensing chips cost only several dollars apiece, could become a staple of cellular telephones and other hand- held devices and might bring the cost of a consumer video camera below $100.
And the contest is not only between film and digital sensors, but between two kinds of chip-making techniques. Foveon's planned announcement, coming on the heels of Kodak's, suggests a sharpening battle between the two competing manufacturing technologies at the heart of a billion- dollar market for digital photographic sensors.
The Foveon chip is based on a low- cost semiconductor industry technology known as Complementary Metal-Oxide Semiconductor, or CMOS (pronounced SEE-moss). The Kodak chip's sensor is based on a more expensive manufacturing technology known as Charged Coupled Device, or C.C.D. imaging.
C.C.D.'s now dominate the digital- imaging industry, but compared with CMOS devices, they require production and assembly of several chips and related components to combine the sensing and computing tasks that can be performed by a single CMOS chip.
Both companies' achievements have startled industry experts because the new devices move far beyond the current industry standards for CMOS and for C.C.D. cameras, which until now have been able to achieve resolutions of 6 million pixels a square inch. The Foveon and Kodak sensors can pack 16.8 million pixels into a square inch.
"If you asked someone if this was achievable in either technology two weeks ago, they would have said it was impossible," said Michael Berger, an industry analyst at Frost & Sullivan, a market research firm in San Antonio.
The Foveon announcement is seen as a personal triumph for Mr. Mead, 66, who is regarded by many executives and engineers as the father of the modern semiconductor industry.
"Carver has tapped into something enormous," said Alexis Gerard, president of Future Image Inc., a digital-imaging research and consulting firm in San Mateo, Calif. "When digital imaging and the telecommunications infrastructure converge, they will enable a shift from a text-based communication model to an image-based model."
Throughout his career Mr. Mead has explored the idea of duplicating the human senses, including vision and hearing, by using silicon-based chip technologies. Several years ago, his earlier work led to the development of a smaller and more effective hearing aid now sold by Sonic Innovations of Salt Lake City.
Complex chips first became feasible in the 1970's after Mr. Mead, at the time a professor at the California Institute of Technology, teamed up with a Xerox computer scientist named Lynn Conway to invent a technique for placing thousands of transistors on a chip a technique known as Very Large Scale Integrated Circuit, or VLSI design.
Today, CMOS-based manufacturing which is used to carry out VLSI design is employed by virtually all microprocessor and memory makers. As a result, it has become extremely cost-efficient and can yield circuitry with more transistors and lower power requirements than most competing technologies.
Yet despite their promise, CMOS- based sensors have until now had just a tiny impact on the overall market for digital imaging because they have been unable to achieve the resolution and clarity of C.C.D. sensors. The global market in 1999 for C.C.D. sensors was $959 million, compared with only $14.2 million for CMOS sensors, according to Frost & Sullivan.
But even before Foveon's latest achievement, CMOS was gaining ground. Not only have companies including Kodak and Polaroid begun to offer inexpensive, low-resolution CMOS-based cameras, but telecommunications giants like Nokia of Finland and NTT DoCoMo of Japan are planning to include inexpensive CMOS sensors in millions of their next-generation cellular phones.
Foveon's contribution has been to improve the quality of CMOS images by continuing to put more computer processing power behind the task of capturing the digital image. The new 16.8 million pixel device has seven active transistors for each pixel. The benefits include less interference, better focusing and more precise exposure times. "When the pixels get smarter," Mr. Mead said, "that translates into better image quality."
Foveon's principal investor and the company's technology partner is National Semiconductor, a big Silicon Valley chip maker.
National Semiconductor's manufacturing plant in Santa Clara is capable of etching chip circuitry only 0.18 micron wide a microscopic fineness that few other chip makers can equal. By contrast, most current low-cost CMOS sensors are made with circuitry of 0.35 or 0.5 microns, which allows for millions fewer transistors per chip.
National Semiconductor executives said the company was planning to take the technology that Foveon had developed for the priciest reaches of the professional photography market and make it economical enough for some new consumer electronics.
"National's interest is not in thousands of cameras a year but in hundreds of millions of cameras a year," said Brian L. Halla, the company's president and chief executive. "We could make the world's highest-resolution throwaway digital camera and sell it for the price of a similar Kodak system."
Foveon officials said they would demonstrate the new 16.8 million pixel sensor for the first time today. The sensor, which for now captures images in black and white, has almost 70 million transistors or about two and a half times the number of transistors used by a Pentium III microprocessor chip for computers. Foveon says it expects the new sensor to be on the market within a year.
Currently, Foveon sells a high-end camera using an earlier version of its sensor that has a resolution of 2048 x 2048 pixels, or 4.19 million in all. That camera uses three separate sensors and a prism array to separate color information. But Mr. Halla said the company was also working on a technique that would permit a single chip to capture precise color information.
Despite the advances now being made, Mr. Mead acknowledged that digital-image sensors are still a long way from matching the skills of the human eye.
An eye is movable, which enables it to scan various parts of an image and then allow the brain to compose a single, larger image. The eye is also remarkably diverse: elements that have high resolution are clustered at the center of the field of vision, while sensing elements that function well at low light levels are around the periphery of the eye, giving human vision a great flexibility of range in varying light conditions that no artificial imaging system can yet match.
Mr. Mead said that because of fundamental size limits in the wavelengths of light, it is unlikely that future digital sensors will gain much additional resolution. Instead, shrinking semiconductor circuit sizes will make it possible for companies like Foveon to add more and more intelligence to their digital-imaging systems, perhaps simulating more of the image-enhancement functions of the human brain.