In dark environments, it is often difficult to get a clear, high-quality image. To address this problem, a method was developed for spherically curving the flat image sensors found in digital cameras. The curved sensors could be used to create better cameras for surveillance, head-mounted displays, and advancements in autonomous vehicle navigation.

Researchers developed a way to create spherically curved image sensors by three-dimensionally bending off-the-shelf image sensors. When incorporated into prototype cameras, the curved sensors produced greatly improved image quality compared to high-end commercial cameras. (Image: Microsoft Research)

Most of today's cameras use lenses made of multiple optical elements that correct for various optical errors, or aberrations, and that also manipulate the image so that it can be detected by a flat sensor. Using a curved, rather than flat, image sensor means the optical elements have to do less work to correct and flatten the image, making it possible to use fewer optical elements. This not only translates to smaller, faster, and less expensive lenses, but also makes it easier to improve other properties of the optical components.

When using curved sensors, it is possible to correct aberrations in a much more efficient way, making it easier to create very-wide-angle lenses that produce sharp images across the entire field of view, or to create fast lenses that produce better images in low light.

The new method can create image sensors with three times more spherical curvature than reported previously. Adding spherical curvature to an off-the-shelf image sensor can be done for a reasonable cost, and in a way that shows significant benefits, including the ability to take pictures under very low light, be very small, and produce extremely sharp pictures.

To make curved sensors, individual sensors cut from a thinned CMOS image-sensor wafer were placed into custom-made molds, and then pneumatic pressure was used to push each sensor down into the mold. Other attempts at curving a sensor typically have involved gluing the edges down and trying to push on the center of the sensor; however, this creates points of high stress that cause the sensor to shatter before it reaches the target level of curvature.

Significantly more curvature was obtained from the sensors by letting them float freely during the bending process, which allowed stresses to dissipate gradually. A specially shaped mold was used that very slowly builds stress around the chip's edges as it is pressed into the mold.

Tests showed that curving the sensors did not change any of their electrical or imaging characteristics. When used in a prototype camera with a specially designed f/1.2 lens, a curved sensor exhibited a resolution more than double that of a high-end SLR camera with a similar lens. Toward the edges of the image, the curved sensor was about five times sharper than the SLR camera. Although most cameras exhibit decreased light detection around the corners of the imaging sensor, the curved sensors lost almost no light. This was a significant improvement compared to the decrease of around 90 percent measured for the commercial SLR camera.

The prototype camera is about the size of a small consumer camera; how-ever, the lenses could be made small enough for mobile phones and tablets. It should also be possible to build machines that could mass-produce these curved sensors, allowing the additional processing to be incorporated into existing sensor manufacturing in a way that would amortize well in volume production.

The researchers are working to see if further improvements might produce sensors with even more curvature. They also want to experiment with curving sensors that operate in infrared wave-lengths, which could be useful for telescopes, three-dimensional (3D) spatial mapping, biometric authentication, and various other scientific applications.

For more information, contact Brian Guenter at This email address is being protected from spambots. You need JavaScript enabled to view it..


Tech Briefs Magazine

This article first appeared in the August, 2017 issue of Tech Briefs Magazine.

Read more articles from this issue here.

Read more articles from the archives here.