29 May 2008
Collecting light through the backside of a CMOS sensor allows OmniVision to extend its pixel roadmap down to 0.9 microns.
OminVision and its partner Taiwan Semiconductor Manufacturing Corporation (TSMC) have unveiled a CMOS architecture that collects light through the backside of the sensor. This radically different approach to traditional CMOS image sensor design will allow OmniVision to extend its pixel roadmap down to 0.9 pixels and continue to miniaturize digital imaging technology.
"Although backside illumination concepts have been studied for over 20 years, up until now nobody has been able to successfully develop the process for commercial, high-volume CMOS sensor manufacturing," said Ken Chen, senior director of Mainstream Technology Marketing at TSMC. "We have delivered a truly advanced technology that defines the future of digital imaging."
According to OmniVision, the BSI architecture delivers a number of performance enhancements over front side illumination (FSI). Improvements include increased sensitivity per unit area, increased quantum efficiency, reduced cross talk and photo response non-uniformity. BSI technology also permits a much larger aperture size, which allows for lower f stops.
In the BSI approach, OmniVision's traditional CameraChip sensor is turned upside down so that it collects light through what was previously the backside of the sensor, the silicon substrate. This approach differs from conventional FSI image sensors, where the amount of light reaching the photosensitive area is limited by the multiple metal and dielectric layers required to enable the sensor to convert photons into electrons.
The front side illumination (FSI) approach can block or deflect light from reaching the pixel, ultimately reducing the fill factor and causing additional problems, such as cross talk between pixels. BSI reverses the arrangement of layers so that the metal and dielectric layers reside below the sensor array, providing the most direct path for light to travel into the pixel. This approach optimizes light absorption, enabling OmniVision to build a 1.4 µm BSI pixel that it says surpasses all the performance metrics of 1.4 µm and most 1.75 µm FSI pixels.
Since light directly strikes the silicon, the fill factor of the image sensor is improved, which provides enhanced low-light sensitivity. A much higher chief ray angle enables shorter lens heights which in turn allows for thinner camera modules ideal for use in the next generation of ultrathin mobile phones.
According to Howard Rhodes, vice-president of Process Engineering at OmniVision, moving FSI pixel architectures down to 1.4 µm and below is difficult because metal lines and transistors are driving the aperture of the pixel close to the wavelength of light – its physical limit.
"To overcome this with traditional FSI pixel technology would require a migration to 65 nm copper process technologies, which would significantly increase the complexity and cost of manufacturing," he said. "Because it allows for more than three layers of metal, BSI achieves significant manufacturing benefits without moving to smaller process nodes. This means routing can be simplified and die sizes can be smaller than in FSI sensors."
OmniVision is currently demonstrating an 8 MPixel, OmniBSI CameraChip sensor, and expects to start sampling first products before the end of June.
|© 2023 SPIE Europe||