New techs raise bar for camera size, sensitivity

Camera phone and consumer digital camera suppliers may have gotten a shot in the arm in June from a pair of technology developments, announced by Tessera Inc. and Eastman Kodak Co.

Tessera has unveiled a wafer-scale camera technology. Meanwhile, Kodak has developed new color filter patterns based on its own, well-established Bayer pattern. Both promise to meet consumers' insatiable appetite for smaller-form-factor and higher-performance cameras.

Tessera's licensable technology for aligning and bonding lenses at the wafer level addresses a nagging issue for many camera module vendors: Optics, which don't scale well, have been a barrier against further shrinks in mobile-phone camera modules.

Kodak's color filter patterns address another problem that has long dogged the digital imaging industry: trade-offs between resolution and sensitivity.

Kodak claims its color filter patterns are designed to more than double the light sensitivity of the CMOS or CCD image sensors used in camera phones or digital still cameras.

The color filter patterns build on Kodak's widely used Bayer pattern—an arrangement of red, green and blue pixels—by adding a fourth pixel to the current RGB arrangement on the sensor. The fourth pixel "has no pigment on top," said Michael DeLuca, the market segment manager for image sensor solutions at Kodak. Such "transparent" pixels, which are sensitive to all visible wavelengths, are designed to absorb light.

DeLuca called the invention "the next milestone" in digital photography. He likened its significance to ISO 400 color films, introduced in the mid-1980s, which have enabled consumers to take pictures under low-light conditions.

Under the new approach, the fourth pixel, called a panchromatic pixel, allows a black-and-white image "to be detected with high sensitivity," according to Kodak. The RGB pixels present on the sensor are then used to collect color information, which is combined with the information from the panchromatic pixel to generate the final image.

Digital photos shot by image sensors without Kodak's new color filter pattern (top) and with the technology. For both, shutter was 1/10s using ISO 100.

Observers described the panchromatic-pixel concept as both simple and elegant. "The technique is admirably simple—open the window to let in more light. It's almost inconceivable that nobody else thought of or acted on this idea until now," said Tony Henning, editor of The Mobile Imaging Report for Future Image Inc.

Megapixel warfare
In this era of megapixel warfare among vendors of ever-shrinking camera phones and consumer digital still cameras (DSC), image sensor companies have scrambled to pack more pixels into sensors. The move to higher resolution in the same form factor, however, requires smaller pixels with a reduced ability to gather light.

"Much of the effort to improve sensor sensitivity has been to boost the gain in the signal from the pixels, but this introduces noise/grain in the images," said Ed Lee, director of consumer services and digital photography trends service at InfoTrends Inc. "So the trade-off has been sensitivity vs. signal noise."

Noting that such a trade-off is not new for the digital imaging industry, some major manufacturer and designers of CMOS image sensors "claim at least a degree of success in compensating for smaller pixel sizes and reduced light sensitivity," said Henning.

There are several ways to tackle the issue of pixel size vs. light sensitivity issue, said Tessera CTO Mike Feldman.

One approach is to use an entirely different image sensor, such as Foveon Inc.'s direct image sensors. The company uses three layers of pixels embedded in silicon. The layers are positioned to take advantage of the silicon's absorption of different wavelengths of light at different depths. Each pixel stack, in theory, directly records all of the light at each point in the image.

But while Foveon's sensors have intrigued many players in the digital imaging industry, for various reasons, they haven't exactly taken the world by storm.

A second approach is to create a larger pixel size by using one camera that consists of an array of thinner cameras.

The third option is to develop advanced signal processing to improve sensitivity, said Feldman.

But the most logical way to improve sensitivity is to see "what filters can do," Feldman said.

On specific product-level efforts to improve sensitivity, Henning observed that some reduce the number of transistors per pixel, leaving more room for the photodiode, or they reduce the size of the metal layers to make more room for the photodiode. "Others reduce the distance between the photodiode and the microlenses above them," he said. Still others actually use those metal grids to help direct the light to the photodiode, he added.

Comparison between the Bayer pattern and Kodak's new high-sensitivity pattern. Kodak's new color filter pattern introduces panchromatic pixels to standard RGB array. Panchromatic pixels are there to detect all visible wavelength to improve light sensitivity.

A different view
The photo giant's refinement of its color filter pattern technology has the potential to set it apart from its competitors because the company "is looking at it from a different direction," said Jon Erensen, senior research analyst at Gartner. Other image sensor companies may also be experimenting with color filters based on different arrays of pixels, Erensen said, but "I haven't heard" of anyone implementing it.

This may be exactly what Kodak needs for its own CMOS image sensors, which are seeing competition from such formidable rivals as Micron Technology Inc. and OmniVision Technologies Inc.. The company began fielding its first CMOS image sensors for consumer DSCs and camera phones two years ago. But compared with its presence in the high-end professional DSC segment, it was not very successful, observed Erensen.

Kodak's color filter technology can be used on any image sensor, whether CMOS or CCD. The company's primary target, however, is its homegrown CMOS sensors designed for consumers.

The first Kodak sensor deploying this technology is expected to be available for sampling in Q1 of 2008, according to DeLuca.

Technical challenges
To reconstruct a full-color image, Kodak has also developed new software algorithms specifically designed to work with the raw data generated from the image sensors that use the new color filter patterns. In theory, the algorithms use the more-sensitive panchromatic pixels to act as the luminance channel of the final image and derive chrominance information from the color pixels on the sensor.

There is one concern, however. Several analysts said that additional overhead might be needed for processing purposes, as the device compiles and calculates the image information derived from image sensors based on the new color filter patterns.

"It does introduce an extra layer of complexity or artificiality in the algorithms necessary to reconstruct the RGB color at every pixel," Henning said.

But "the only thing that matters," Henning added, "is the quality of the pictures consumers get when using camera modules incorporating all the latest high-tech developments."

- Junko Yoshida
EE Times