Rambus is developing a so-called pico-camera, a lens-free device significantly cheaper and smaller than today's tiniest imagers. The technique still requires significant computation to deliver even a blurry image, however the company sees an initial opportunity to enter the market as a thermal camera.
The company created an 80 micron diffractive grating from which it has constructed a poor but visible image of the Mona Lisa (below). The work required computation on the order of 350 million multiple-accumulates
"We might cut that by an order of magnitude, but you are still talking serious number crunching -- but who cares, there's plenty of compute around," said Patrick Gill, a senior research scientist in Rambus Labs, speaking at the Trillion Sensor Summit here.
Today Medigus Ltd. makes some of the world's smallest imagers, 1 mm2 devices that cost about $1,000 and are used in endoscopy cameras. Rambus foresees much smaller imagers that could be disposable, in part by doing away with a focusing lens.
"This is where the vision of snails and insects lives -- there's useful information here that's not accessible with the focusing cameras we have built," Gill said.
The technique is based on a capturing light on diffractive grating. The image is transferred to a diode array digitized and manipulated essentially using a set of linear equations and techniques based on Fourier transforms.
The grating requires just 100 microns of silicon area, making it suitable to put on a typical microcontroller. Such imagers could be used as disposable devices in a variety of apps for hazardous or medical environments. They could read bar codes, recognize faces, track objects, or estimate motion, Gill said.
Rambus is considering a reference design using the technology for thermal cameras. Thermal cameras are increasing in volume in automotive and other apps, but can only scale down so far in costs due to their need for lenses.
The image (top center) is the best Rambus delivered to date from an 80 micron device, but the company expects to double resolution with advanced math.
The company got its first experimental results back in August. It aims to publish results soon.
This article was originally published on EE Times.