Single CMOS Image Sensor Sees in 3-D

An email arrived from Canesta, a company that I hadn’t heard of, so I went to the Canesta Web site where I learned the company makes 3-D CMOS image sensors.

You can buy or design 3-D vision systems with off-the shelf cameras, but most of these systems use off-axis laser beams or several cameras to grab images that intensive software processing turns into a 3-D image. That means they require careful calibration and alignment. The Canesta sensors bypass those requirements, although it need a plastic or glass lens to focus an image on the sensor’s active area that “measures” 176 by 132 pixels. Each pixel comprises two detectors, one samples in phase with pulses of IR laser light and one samples out of phase. The difference in light accumulated at each detector represents a distance based on the duration of the IR light pulses generated at 44 MHz.

Canesta’s white paper, “Introduction to 3D Vision in CMOS,” described the sensor operation in detail. Find it at: www.canesta.com/assets/pdf/technicalpapers/Canesta101.pdf. For general information, go to the whitepaper, “3D Vision Enables Everyday Devices to ‘See’” at: www.canesta.com/assets/pdf/technicalpapers/Why3d.pdf. You can find other papers and application notes on the company’s Web site under the “Products and Technology” in the “Technical Papers” section.

I talked with Jim Spare, Canesta’s president and CEO, who explained that the company aims primarily at high-volume automotive and consumer applications, at least for now. The company will pursue other markets as high-volume applications arise. So far, Hitachi has developed a prototype TV controlled by user gestures such as waving to turn it on and using a circular hand motion to change channels. You can watch a short video of the TV’s operation at: www.youtube.com/watch?v=mDA8g_nIAU8.

The sensor produces a 12-bit value for active light data and an 8-bit value for brightness data. Ambient light doesn’t affect the sensor’s operation. A 16-bit parallel digital interface and DMA controller transfer pixel information to a host device for processing. But, this software does not have to construct a 3-D image; the pixel information provides distance values.

I had hoped to have some images of chips or cameras to show, but so far the company doesn’t have much “public-relations-type” information to share.  Canesta recently got an infusion of $16M from Quanta Computer and SMSC, so you might see a “breakout” into other markets and and more forthcoming information about how designers outside “big” markets can take advantage of the company’s neat 3-D vision chips.

Canesta does NOT yet have development kits, but you can sign up for an alert as the company provides more information and starts to offer engineering and evaluation tools. Go to: canesta.com/developers/request-more-information. The sensor looks like an interesting product and I bet engineers will find cool applications for it. –Jon Titus