Tuesday, August 9, 2005


Interesting interview of George Gilder over at AlwaysOn:

George Gilder: I was writing a larger book about Carver Mead, and I got embroiled in this story, and it seemed to be an exciting story in itself. It also was a meaningful entrepreneurial story and scientific story. So I separated it out from the larger narrative that was about physics and focused on Foveon, which is a company that Carver Mead launched in 1996 to revolutionize photography.

What makes his work with the company such a big idea?

Gilder: It put analog technology on the digital Moore's Law learning curve. Until Carver began these experiments in the early '90s, analog devices were on a much larger scale than digital devices. They were five or six times larger, and that is the transistors they used. They were not capable of the kinds of advances that Moore's Law and part of the digital technology [allowed]. Carver used the same CMOS processes, the same plain-vanilla silicon processes that are used to make every Pentium microprocessor to make analog devices, and these analog devices could run sub-threshold. They run before the device switches; it functions as an almost perfect linear analog device. Making these small analog transistors made it possible to do a pixel that could produce a very high-resolution image. So they launched the project of making an analog pixel that could capture all the colors at every point in an image, unlike all other digital cameras. All digital cameras use the Bayer mosaic technique. Essentially they are light detectors. They only detect the intensity of the light, and in order to create colors they filter, they impose a filter, and the filter filters one color at every pixel—one primary color, green, blue, or red at every pixel. And then the actual colors of the image are calculated digitally.

The difference of the Foveon device that took a long time to perfect was that it collects all three colors in analog form at every pixel. And it derives from the well-known but little-used frequency-dependent penetration of light into silicon. The different colors penetrate different depths in the silicon. So you can collect the blue light near the surface, the green light a micron or so down, and the red light two microns down roughly—and thus can create an analog device that corresponds to the colors that human eyes detect.
(full post)

No comments: