Foveon article in June issue of Forbes:
forbes.com
An upstart led by a silicon wizard hopes to sell digital cameras to an untapped market: the pros.
Sharper image
By Philip E. Ross
DIGITAL CAMERAS would seem to be the last product a startup would want to make. They already are on the market, in versions priced at $300 to $40,000, and no less than Intel and Kodak have joined forces to work on the next generation. Kodak pulled in well over $1 billion from digital products last year.
How in the world can little Foveon, a Santa Clara, Calif.-based piker with not a dime in revenue, hope to compete? By aiming a superior product at a niche the giants have poorly served: the nation's 60,000 professional portrait photographers. Leading the effort is a brilliant mind-Carver Mead.
A professor at California Institute of Technology and winner of the $500,000 1999 Lemelson Prize, Mead made his name in semiconductor research in the 1960s. He devised a gallium-arsenide transistor still in use today, correctly theorized that miniaturization would go much further than anyone thought and laid down the design rules for etching millions of components onto a chip.
In the 1980s he looked at the signal processing of natural computing systems such as the cochlea of the ear and the retina of the eye. That led to work on an artificial retina, which prompted him to pursue a better digital camera. He founded Foveon in 1997, naming it after the retina's center of vision. (The company isn't related to an identically named firm that tracks Internet data.)
Mead lined up backing from National Semiconductor (which owns 45% of Foveon) and Synaptics (30%) and went to work. Sales of digital cameras this year are expected to hit 1.4 million units and approach $600 million in revenue, but those products are cheap and crude. Professional digital cameras sell for up to $30,000 but have technical flaws, especially in making prints 16-by-20 inches or larger.
The flaws stem from the lack of enough data points for the camera to process an image properly. In digital cameras, in place of light-sensitive film, light focuses on a chip (generally made by Sony or Philips) that consists of microscopic, light-sensitive cells called charge-coupled devices. These cells lie in orderly arrays, and each has a filter to detect blue, green or red while remaining blind to the other two colors. That means the array doesn't see every color at every point; to fill in the gaps, a computer guesses at the missing data.
"The big thing that goes wrong with essentially all the digital instruments out there is they don't have information about every color at every location," Mead says. "Because they're guessing, they get fooled."
When computers connect the dots, the pixels can end up looking like twinkles of different colors of light where a clean edge should be. The twinkly Christmas-light effect shows up when the camera must capture fine lines such as strands of hair. A more serious defect is moiré patterning-those wriggling whorls that jump off striped shirts on television. It comes from viewing one grid through another, as when you look through one screen door into a second screen.
"We sense color at every point," Mead says. "There's no guesswork."
Meade won't say exactly how Foveon avoids moiré madness. He does allow that the light-sensitive cells in his camera aren't common charge-coupled devices and are controlled not by software but by analog elements embedded into a specially designed chip. "We sense color at every point; there's no guesswork," he says.
Dean Collins, a San Diego photographer, used the Foveon camera and tried and failed to provoke moiré patterns; he still can't figure out the secret of how the new system works. "They said if they told me how they do it, they'd have to kill me."
Secrecy is important, Mead says. While Foveon is applying for about 40 patents, they can do only so much to fend off Intel and Kodak. Intel declines to comment on Foveon's claims, and Kodak says it already serves the portrait market with a solid digital product.
At the heart of Foveon's better mousetrap is a custom-designed set of three chips. Also at work is a beam-splitter that feeds light through three color filters to the three chips, a trick first used in video cameras. (Other systems use one chip to take three successive pictures, a neat way around the pixel shortage, but one that works only for still-life pictures.)
Foveon also hopes to win converts based on ease of use. It takes only three minutes for Rudolph Guttosch, Foveon's marketing director, to click the silicon camera trained on a subject, export the digital image to a photo processing computer a few feet away, fool with the color and other metrics and print it all out. Every stray hair on the subject's head shows up clearly, unmarred by distortions.
Richard Lyon, chief scientist of Foveon, credits that smooth work to system integration. "Instead of retrofitting a film camera for digital chips, we rethought the whole system and redesigned every part: optics, sensors, hardware, software, user interface."
The Foveon viewfinder shows, in advance, roughly what a shot will look like. It appears as a grainy image in a window on the screen of the laptop that's bolted to the lens and chip assembly. A neighboring window displays the last photo you snapped, with any overexposed regions highlighted by software. Such errors can be fixed by reshooting, while the subject is still in the studio. When you're done, you have what you want (instead of "what you see is what you get").
Mead tapped a 37-year veteran of Kodak, Ray DeMoulin, to be president and chief executive; he had run Kodak's professional photography business. The new digital system recently went into beta-testing among portrait photographers to find ways to make it even easier to master and to round up testimonials. A product should hit the market by year-end in the $10,000-to-$30,000 range, the practical limit for high-end gear. Mead says he won't take Foveon public until it shows a solid profit.
But don't hold your breath for a consumer version. "We don't ever want to sell $100 cameras," Mead says. "It's true that as digital technology evolves, you get more performance per dollar-but that doesn't necessarily mean you want to make things cheaper." |
