Five CMOS Camera Developments to Watch in 2014

1. CMOS Sensors Will Go Big. Real Big.

Press coverage started the year with Hasselblad’s “announcement of an announcement” ahead of their new medium format camera sporting a 50MP CMOS Sensor. It’s expected that Phase One will also launch a new camera this year built around the same sensor. Now, at 2 fps, this sensor won’t be driving industrial vision applications very soon, but the demand for higher resolution is constant, and more pixels can generate new possibilities for efficient imaging workflows, particularly in geo-mapping and earth imaging.

Now that CMOS has leaped into a new league (challenging CCD’s stronghold in the medium format camera market), it’s easy to imagine new machine vision cameras that will leverage the ambition of this sensor, if not the exact technology. There are many ways to make a big sensor. How will it make the most sense, and when might “big” be big enough?

2. CCD + CMOS: The Best of Both Worlds?

Also in the realm of “announcements of announcements” is Belgian researcher IMEC’s announcement of a hybridized TDI sensor design. They claim to have developed a ‘CMOS Image Sensor for High Performance Applications’ prototype that is based on some of their CCD technology. The prototype combines a light-sensitive, CCD-based TDI pixel array with CMOS readout electronics bolted onto it. By integrating the two primary imaging technologies, IMEC hopes to successfully combine the best of both worlds. If it works, the result could be CCD’s low-noise performance ideal for TDI imaging, but with the low power and fast readouts that only CMOS can deliver. Hybrid sensors in and of themselves aren’t new, but taking this approach to high-performance TDI imaging is an interesting workaround to the fact that CMOS has yet to succeed where CCD has dominated for so long.

3. Accelerating Speeds and Feeds

As CMOS sensor performance continues to improve, bandwidth at the other end of the camera will continue to be a challenge. USB 3.0 is promising, and standards such as Camera Link HS and CoaXPress have been developed to solve exactly this problem, but sensor and camera design keep the pressure on. The megapixel count will continue to rise (see above), so cameras will soon be able to saturate the bandwidth of even these recent interface standards.

There is literally, light at the end of the tunnel, though. Fiber optics offer the possibility of data throughput that can keep up with your camera, and there is already progress on what a standard for fiber optic data throughput might look like.

4. Better Low-Light Performance

The history of CMOS sensors (PDF) is one of constant advancement. But, in the race to deliver higher and higher resolution, pixels have gotten smaller (and thus have less light-receiving photodiode area). This has led to compromises in pixel performance degradation due to declining signal-to-noise ratio (SNR) and well capacity.

To cope with loss of resolution and pixel size reduction, new processes and design innovations have found ways to overcome the limitation of conventional pixel performance. Instead, expect sensor makers to continue to push the adoption of new pixel technologies that can bring back the balance of performance, even with smaller pixels.

Use of features such as pinned photodiode technology (PPD) and optimized implantation techniques can reduce the dark current and manage the appearance of “hot pixels”—key contributors to noise and lag in an image. With a lower overall noise floor, new imagers can be used with less illumination at faster frame rates and still achieve the same image SNR as older designs.

Back-side illumination (BSI) is almost a mainstream technology. Although it’s been in use in CCDs for a long time, bringing it to CMOS sensors and their crowded pixels has led to a dramatic increase of sensitivity. BSI will continue to allow pixels to get smaller and resolution to go up, without a lower penalty for SNR.

5. More Will Be Happening On-Camera

One way to get around the data throughput limitations and to better manage the amount of imaging data created is to task the camera with more work. By embedding pre-processing inside the camera, (e.g. auto-brightness, HDR, multiple region-of-interest, flat field and pixel correction) camera manufacturers are helping to lower system costs and free up CPUs to manage even more tasks with potentially more cameras.

Feature rich cameras that can make region-of-interest selections and even compress images can lighten the data load on the rest of the system—you still get the imaging you need, but with much less demand on the CPU and frame grabber(s).

On-camera processing can also ensure that you get the image exactly as you want it. Auto-brightness, HDR, windowing capability and ability to change camera aspect ratio are examples of how camera design, in conjunction with a CMOS image sensor, can provide additional capabilities to an end user. Real time embedded processing in the camera can also compensate for non-idealities common to CMOS sensors with features such as pixel response non-uniformity correction. The result can be a better image, faster.

Putting Pressure on the Rest of the System

Sensor technology and camera innovations will continue to advance at a rapid pace. They face the fewest fundamental limitations. In contrast, dealing with the output of these cameras — data transfer and storage – will continue to be perennial bottlenecks. While some of the developments listed here are intended to reduce this pressure, interface, frame grabber, hard drive and vision system designers certainly still have their work cut out for them.