Balancing CCD & CMOS Technologies in Industrial Imaging

ON Semiconductor’s Michael DeLuca discusses CCD and CMOS imaging technologies, and the benefits of each.

The broad adoption of imaging systems in industrial applications continues to expand, driven not only by the development of new image sensor technologies and products, but also by advances in supporting platforms such as computing power and high-speed data interfaces. Today, the use of imaging systems in areas as diverse as assembly line inspection, traffic monitoring/enforcement, surveillance, and medical and scientific imaging is now commonplace, fuelled by advances in image sensor technology that enable improved imaging performance, readout speed, and resolution. With image sensors now designed using both charge couple device (CCD) and complementary metal-oxide semiconductor (CMOS) technologies, it is helpful to review these two platforms to help guide selection of the image sensor best suited to a particular application.

The development of electronic imaging began in the late 1960s with the Nobel Prize winning research of Boyle and Smith to develop the first CCDs. These devices operate by leveraging doped silicon’s intrinsic ability to convert photons into electrons, using the resulting pixel-level charge to measure light intensity. Architecturally, one of the great strengths of this design is its simplicity - the entire pixel area can be used to detect photons and store charge, providing a maximum signal level that enables high dynamic range. This same pixel area is used to transport the charge to a limited number of outputs, where the charge is converted to a voltage. Over time, this architecture has been refined to include interline transfer CCD designs, which incorporate an electronic shutter at the pixel level that removes the need for a mechanical shutter in the camera design. Today, CCDs are manufactured using a customized semiconductor process that is heavily optimized for imaging use, and require external circuitry to convert analogue output voltages into a digital signal for subsequent processing. In general, CCDs are typically characterized by having highly effective electronic shutter capabilities, a wide dynamic range, and superior image uniformity.

In contrast, CMOS image sensor designs were first leveraged from processes developed for the manufacture of mainstream semiconductor devices, such as those used in logic chips, microprocessors, and memory modules. This presents a great advantage, as digital processing capabilities can be incorporated directly into the chip to enhance the image sensor’s functionality. Rather than converting charge to voltage at a limited number of outputs on the sensor like a CCD, CMOS image sensors have transistors placed within each pixel (or each group of pixels) to perform charge to voltage conversion. This allows voltage (rather than charge) to propagate throughout the device, enabling much faster and more flexible image readout. In addition, higher-end processing can be incorporated directly on the chip so that, if desired, the output of the image sensor can be a fully processed JPEG image, or even an H.264 video stream.

While CCD image sensors have historically provided better imaging performance than CMOS devices, this gap has closed significantly in recent years, with the image quality available from CMOS image sensors now more than sufficient for adoption in many applications. This can be seen in the latest generation of CMOS devices for industrial imaging, such as ON Semiconductor’s family of PYTHON CMOS image sensors. While some imaging parameters available from the best CCDs may still surpass those available from this family, the image quality of these PYTHON devices makes them suitable for in-line inspection, traffic monitoring/tolling, and motion analysis. This allows the other performance advantages of CMOS technology to move to the forefront, such as access to significantly faster frame rates, lower power consumption, region of interest (ROI) imaging – each of which can be critical in improving throughput and adoption in these applications.

With intrinsic advantages such as these, some have projected the eventual demise of CCD image sensors as CMOS technology continues to advance and eventually eclipses CCD performance in all vectors. While both CCD and CMOS technologies will undoubtedly continue to develop in the future, the underlying architecture of CCDs suggests some areas where they will continue to retain specific performance advantages, making CCDs the preferred technology option for industrial applications requiring the highest imaging performance.

While image uniformity continues to improve as CMOS technology advances, the highest level of performance continues to be found with CCD image sensors. This is a direct outcome of the architecture of these technologies: Though a CMOS device can have thousands of separate amplifiers (one for each column, or even one at each pixel), CCDs can route charge from pixels to a single amplifier, eliminating any amplifier to amplifier variation in the readout of the sensor. High image uniformity can be very important for applications such as medical and scientific imaging or even in critical end-of line inspection, where the quantitative nature of these applications can make having clean, unprocessed images critical. In addition, retaining this uniformity when scaling to high resolutions and large optical formats tends to be easier with CCDs than with CMOS devices.

The analogue nature of a CCD design also allows a CCD camera to be ‘tuned’ for a given end application, optimising specific imaging features for a given end application. For an application such as astrophotography, for example, a camera manufacturer can choose to optimise the full well capacity of the sensor (extending dynamic range) at the expense of anti-blooming protection (which may not be as important for this application). Other scientific imaging applications can also benefit from the very low dark current available from CCDs, where exposure times up to an hour or more can be required for detection of very faint signals.

Because of architectural advantages such as these, ON Semiconductor continues with selected investments in CCD technologies and products today. An important example can be found in the recent announcement of a new CCD technology platform that combines the imaging performance of interline transfer CCDs with the extreme low light sensitivity available from electron multiplication (EMCCD) outputs. This interline transfer EMCCD combination allows a single camera to capture an image where one portion of the scene (such as an alleyway) is illuminated at very low light levels (down to moonlight, or even starlight), while a separate portion is lit with a bright light (street lamp). This capability, enabling a single camera to capture from daylight to starlight, is unique to CCD technology as it leverages the charge multiplication nature of the EMCCD output – a feature that is not available with CMOS devices operating in the voltage domain. The first product to incorporate this technology, a device with 1080p resolution operating at up to 30 fps, is targeted to very low-light applications such as surveillance, scientific imaging, and medical imaging.

These images show the same scene shot by different technologies:

1. EMCCD

2. Interline transfer CCD

3. Interline transfer EMCCD combination

While it is tempting to try to identify a ‘winner’ when comparing CCD and CMOS technologies, this really does a disservice to both, as each technology is unique and provides distinct end user advantages. While products using CMOS technology are clearly becoming widely adopted, there still remain areas where the CCD image sensors retain advantages, making them a better fit than CMOS devices for some applications. Rather than looking to find the best technology, therefore, it is more important to identify the key performance parameters of the particular end-use case being considered, and then map those critical needs against the features and performance available from different products. While in some cases products based on one technology may provide the best match, in others the situation may not be as clear – making it important to work with a company that can represent both technologies in order to obtain an unbiased view. With access to a broad portfolio of products based on both CCD and CMOS technologies, end customers can then identify and select the device truly most appropriate for their given end application – making them the real winner.

ON Semiconductor