Today’s machine vision cameras and sensors come in a multitude of resolutions, optical formats and technologies.

Pictured is the Truesense Imaging 5.5-Micron Interline Transfer CCD Platform of electrically compatible image sensors that differ only in optical size and resolution. Source: Truesense Imaging Inc.

Each of the primary sensor technologies, CCD and CMOS, find homes in various applications and camera makers must choose which best fits the needs of a particular application. CCD sensors have existed for many years and offer enhanced electronic shuttering and improved image quality. CMOS sensors offer high frame rates, but with compromised image quality and less than ideal electronic shuttering. Generally, CMOS sensors and cameras serve the lower end applications such as determining if a lid is present on a container, while CCD sensors and cameras continue to serve the applications that demand the utmost image quality. These high-end applications will continue to be served by high performance CCD sensors and cameras, and they will co-exist with their CMOS counterparts into the future. This article focuses on the trends that have evolved for high performance CCD sensors and cameras.

Several key applications for machine vision have been pushing the envelope for CCD sensor and camera performance. One such application is aerial surveillance, which has an ever-growing requirement for increasing resolution in order to capture finer details over a wider area. Similarly, automated optical inspection (AOI) of flat panel displays continually demands higher resolution to handle growing television sizes and increased adoption of high definition in consumer handheld devices, such as tablets and mobile phones. Further, AOI has a never-ending requirement for faster image capture and processing in order to maximize production throughput.

Unfortunately, increasing sensor resolution tends to slow image capture and processing. Higher resolution means increased pixel count, which increases the amount of image data that a camera must handle in each frame. This data increase, in turn, requires the camera to become faster at reading the sensor and sending information to the host processor in order to avoid reducing the frame rate and impacting production throughput.

Within the family, a camera designed for a 1Mp image sensor would also work with the Imperx 29Mp camera (pictured). Source: IMPERX

Pixels Shrinking

Meeting these demands for higher resolution and faster data throughput has been the driving force in many of the trends in sensor and camera development during the last few years. To increase resolution, for instance, CCD sensor developers have been shrinking pixel geometry in order to increase pixel count while maintaining standard sensor sizes. Shrinking geometry, however, negatively affects the sensitivity, well depth and dynamic range of CCD sensors, thereby degrading image quality.

Sensor vendors have thus had to improve their device designs and fabrication processes to compensate for the effects of shrinking pixel geometry. Truesense Imaging, Inc. (formerly Eastman Kodak Company, Image Sensor Solutions), for example, made several design and process improvements when moving from 7.4 µm pixels to 5.5 µm pixels in their interline CCD products. The combination of these improvements enabled the 45% smaller pixels to maintain sensitivity and dynamic range, while improving the smear rejection ratio to -100 dB, when compared to the larger 7.4 µm pixels.

Along with shrinking pixels to increase resolution for traditional sensor sizes, vendors are now offering larger sensors. These large sensors not only increase total pixel count, they allow development of cameras that are capable of high- resolution imaging with a wider-area field of view.

Another trend in CCD sensor design is an architectural evolution toward multiple simultaneous readout paths. Early CCD sensors had a single shift register with a single output amplifier, transferring out image data one pixel at a time, by rows. As pixel counts grew, however, the time needed to read an image out of the sensor also increased, reducing the frame rate that the sensor could achieve. The next progression was to separate the image array into two sections, each with its own output path, which effectively doubled the achievable frame rates. Sensor architectures are now moving from a two-tap to a four-tap structure, which breaks the image array into quadrants for independent readout (Figure 1), which allows even faster frame rates.

Device and Camera Families Emerge

CCD cameras are finding homes in an increasing array of applications with widely varying needs for speed, resolution, and pricing. To address these diverse needs, sensor vendors have begun creating image sensor families. These families consist of devices with a common architecture and common performance characteristics, while only differing in optical size and pixel count (Figure 2).

These sensor families also offer a common electrical pinout, allowing a single hardware design to support all sensors in the family. A camera designed for a 1Mp image sensor, for example, would also work with the 2Mp, 4Mp, 8Mp, 16Mp and 29Mp sensors within the family. (While sensor size differences force a slightly different physical pin layout, the electrical design remains consistent.) Additionally, the sensors in this family offer an identification pin that identifies the resolution of a particular image sensor, enabling the creation of a camera that can automatically detect the sensor that has been installed and adjust the camera settings to correctly operate that particular image sensor.

The emergence of sensor families has prompted a corresponding trend in camera design toward the creation of camera families. As little as two years ago new sensors appeared at the rate of one or two per year and camera developers crafted unique designs around them. With the advent of sensor families, however, camera developers have had to change their design approach. Now, they are creating designs that will support multiple sensors within a family, sharing common electronics and hardware elements. The family concept of the image sensors has also shortened the time-to-market for new camera introductions. These camera families, such as the Imperx Bobcat series, give users multiple resolution, speed and price options with a consistent mechanical and functional interface.

High frame rates in high resolution CCD sensors are enabled through the use of four-output architectures that allow the data to be read out in parallel, such is the architecture shown for the 29 Mp Truesense Imaging KAI-29050 Image Sensor. Source: Truesense Imaging Inc.

Faster Interfaces Arise

Camera vendors are also responding to the architectural changes in sensor design by increasingly adopting the four-tap readout. This adoption complicates camera design somewhat, however, because of the need to compensate for variations in the sensors’ output amplifiers. Because the amplifiers are independent, they will produce a slightly different mapping from pixel charge level to output data value. The result is a slightly different brightness and contrast in the four image quadrants, which is most visible at the seams between quadrants. Cameras need a mechanism for correcting for these differences.

The adoption of four-tap readouts along with higher resolutions is also forcing a change in the design of camera interfaces to the host system in order to keep up with increasing image data transfer rates. Early digital cameras for vision systems often used the CameraLink interface, with a multi-wire cable that linked camera and frame grabber. To reduce the interface cable’s cost and complexity as well as to extend maximum cable length, camera vendors created the GigEVision interface based on the Gigabit Ethernet standard.

Multi-tap, multi mega-pixel sensors, however, are pushing camera interface requirements to data rates beyond the Gigabit Ethernet capacity, forcing camera developers to develop even faster interfaces. During the last two years two such high-rate interfaces have arisen: HS Link and CoaXpress. Camera vendors are now incorporating these interfaces in their emerging high-performance camera families.

Along with the higher data rate, camera vendors have begun incorporating an additional feature into their interface designs: power over cable. By allowing the interface to carry power to the camera along with supporting control and data transfers, vendors have removed the need for separate power cabling and made cameras easier to install. Virtually every popular camera interface standard now supports a power over cable option.

These trends in CCD sensor and camera design are the result of the continued growth in the applications for high performance machine vision systems and the demands placed on such systems. This growth has changed the way sensors and cameras are developed and resulted in a wide range of options not imaginable as little as two years ago. The range of options will continue to grow with the support of ever faster and higher resolution sensors, combined with the development of higher-rate camera interfaces.V&S

Tech Tips

Improvements in the technology for CCD products allow 45% smaller pixels to maintain sensitivity and dynamic range,

With these improvements, a 5.5 µm will improve the smear rejection ratio to -100 dB, when compared to the larger 7.4 µm pixels.

A camera designed for a 1Mp image sensor will also work with the 2Mp, 4Mp, 8Mp, 16Mp and 29Mp sensors within the family.