While charge-coupled device (CCD) and complementary metal-oxide semiconductor (CMOS) sensors perform similar functions — collecting light (photons) and converting those into charges (electrons) — there are a few fundamental differences between their methods. The biggest difference is in the sensor architecture. While CCD sensors generally employ a majority of the pixel surface for capturing light, CMOS sensors use a portion of that surface for conversion electronics.
Because up to a few years ago the conversion electronics required a significant amount of space, CCD sensors were generally considered to be more light-sensitive. Modern CMOS sensors are now outfitted with space-saving microelectronics. The numerous, comprehensive improvements have brought microlenses directly on the sensors, to bundle the incoming light and direct it straight to the light-sensitive parts of the sensors.
The results are clear: CCD and CMOS sensors are almost identically effective at capturing photons.
CMOS Area and Line Scan Cameras
If you look at two different applications of CMOS technology — area scan and line scan sensors — you’ll find that there’s another difference of major importance: unlike area scan sensors, modules with line scan sensors have only one row of pixels available. This generally provides extra space for the necessary conversion electronics to be located outside the pixel. In practical terms, the fill factor for line scan CMOS sensors totals almost 100%.
High line frequencies are the major strength of CMOS technology. Its structure inherently keeps delays between transfer and processing of image data to a minimum, which in turn enables the very high line frequencies that are essential to industrial applications such as print quality inspections, as well as for medical technology and research. For many of these applications, it is so important to keep the highest possible localized resolution (including sampling rate) that area scan cameras are no longer sufficient to implement the image processing. Only line scan cameras are up to the task.
The current standard for line scan camera applications calls for line rates of 10-20 kHz (fps). Constantly rising demands on throughput, such as where image processing is needed for quality assurance applications, has fed a clear trend toward higher speeds, even in line scan cameras. Higher line frequencies in the range above 40-140 kHz remain rare even today, but in the future, they will become commonplace.
Because the speed limit for line scan cameras based around CCD technology top out around 30 kHz for a 2k line, the trend towards higher speeds in line scan applications means that CMOS technology will eventually be required for line scan cameras as well.
Selection of Interface Technology
The latest generation of CMOS sensors allows for line frequencies of up to 51 kHz via a GigE interface, as well as up to 80 kHz via Camera Link, with resolutions of 2-12k as needed. This makes clear the tremendous potential for this combination of CMOS sensor and high-performance camera.
The use of quicker CMOS technology has also had an impact on various interface technologies. While speed limitations inherent to CCD sensors just a few years ago meant that it was not possible to achieve high line frequencies of 50 kHz or greater at 2k resolution when using Camera Link, the classic line scan standard, the CMOS sensors at work in today’s cameras are just starting to test the tremendous potential and bandwidth available through the GigE interface. More than ever, CMOS line scan cameras with GigE interface are a real alternative to cameras outfitted with a Camera Link interface.
AOI settings that limit the resolution of the camera to 1024 pixels mean that the interface is no longer the limiting factor, with the maximum attainable camera speed corresponding more or less to the maximum read-out speed on the sensor. In many modern line scan applications, the speed potential, the flexibility, and attractive price for the overall concept of a GigE-based line scan solution have proved so convincing that the interface has enjoyed lasting tremendous success, including in the line scan segment.
Low-Voltage, Low-Heat, Little Space
CMOS offers more than just high frame rates. Since all functionality is implemented in the image sensor itself and, unlike CCD technology, CMOS technology works on a charge basis, the space requirements for signal and power sources can be kept to a minimum. There is no need to transport charges. They are immediately available for further processing, which in turn produces significantly lower power consumption. Because a relatively very low level of current flows within the sensor, this has a highly positive impact on the thermal power loss. Only low levels of heat are generated as a result of power dissipation, and this heat can be transported away simply and without the need for extra cooling elements.
This has a positive impact on the size of the camera casing. Even during the engineering phase, the fact that less heat is generated in the interior of the camera means that less infrastructure is needed to draw away that heat. The result is a more compact camera design.
Industrial applications for inspection of fast-moving piece goods typically involve a variety of cameras arranged next to one another to cover the inspection surface with the greatest possible precision at the maximum frame rate; these are prerequisites for the quality inspection to run at the greatest possible resolution. Where numerous cameras are used, the benefits of a space-saving solution multiply with each unit that is added. Put simply, the smaller the camera housing, the simpler it is to integrate into the overall system. The trend toward miniaturization is consistent and has extended toward pixel sizes as well: pixels of up to 1.5 m are no longer a rarity on CMOS area scan sensors, while pixel sizes of 3.5 and 7 m are standard on line scan cameras.
Outstanding Image Quality
The low power consumption inherent to the CMOS technology impacts more than just the compact design of the camera — it also impacts image quality. The hotter a camera runs when operating, the higher the so-called “read noise” of the image sensor. This is true for both CCD and CMOS sensors. One rule of thumb underscores the benefit of CMOS technology over CCD technology in this regard. For every 6-8 °C that the operating temperature rises within a camera, a doubling of the noise level can be expected. Various values denote and impact the quality of an image sensor. For the high-speed applications typically seen in the line scan camera field, the following characteristics are relevant:
Signal/Noise Ratio: This refers to the point at which the first signal can be detected, i.e. the point at which the noise and signal are equally strong. This sensitivity threshold represents a signal/noise ratio of 1. The maximum signal/noise ratio corresponds to the square root of the full well capacity. Modern CMOS sensors have a slight advantage over CCD sensors insofar as the high speeds (line frequencies) in modern line scan camera applications have led system engineers to continually push the envelope of the possible in terms of sufficient and homogeneous lighting.
Very Low Dark Noise: It is here that an additional benefit of CMOS technology emerges: its extremely low dark noise values. This means a camera can generate signals in dark environments that leave other cameras perceiving only noise. Imagine that a golf ball represents the signal, while the ground it lies on represents the read noise. If the ball is on the green, it’s easy to see it. Even out on the fairway, it’s still possible to detect the ball. Out in the rough, though, the ball disappears from view. It follows that these cameras can also produce very good results even in dark environments and when working with standard illumination (LED). With CMOS, expensive high-end lighting solutions are no longer mandatory. One frequent remark among users of line scan cameras is that the vast majority of applications do not have the truly necessary levels of lighting available. This criterion makes CMOS technologies of value for line scan camera applications.
Full Well Capacity and Dynamic Range: The full well capacity describes the maximum number of electrons that a pixel can absorb. The ratio between the full well capacity and the sensitivity threshold produces the dynamic range, measured in dB. The dynamic range describes the ratio between the largest and smallest signal that can be detected as an image outside the noise. At the same time, it describes a camera’s ability to produce an image that contains both high levels and low levels of light, yet has minimal noise and minimal errors. Although for CMOS sensors the number of electrons per pixel is often artificially capped, they typically have a higher full well capacity than CCD sensors.
When developing modern CMOS sensors, manufacturers are endeavoring to establish benchmarks that are more and more suitable for the human eye. The logarithmic characteristics for the cameras produced in this way can achieve a dynamic range of over 100 dB. This procedure has the effect that the darker area of an image features greater differentiation than in the light areas.
In the area of line scan sensors, this characteristic has been further adapted to meet the needs of various applications. The field places a premium on the optimal combination of high speeds with largely standardized lighting conditions. The result is that a pixel does not need large full well capacity, since it wouldn‘t be exhausted anyway. In today’s sensors, a high dynamic range is also achieved on an application-to-application basis through an extremely low dark noise value together with sufficient full well capacity.
Sensitivity and Light: Another fundamental difference between the two sensor technologies comes in the area of sensitivity thresholds. For CCD sensors, the maximum lies at 550 nm, which is exactly within the most sensitive area of the human eye. For CMOS sensors, however, this falls in the red spectral range, between 650-700 nm. This higher sensitivity to light outside the spectrum visible to the human eye is highly beneficial for many ultrasensitive applications, including for industry, medicine, and foodstuff inspections.
Limited light represents a challenge for almost all line scan camera applications. Because the darker areas are offered with greater differentiation, applications with CMOS sensors require sufficient light. Sensors, such as those with two active lines, provide a valuable service by further enhancing the benefits of strong sensitivity and a strong signal/noise ratio. The sensitivity present in the single cell mode is up to four times higher than for conventional line scan cameras. If each object row is scanned twice when in dual line mode, then the signal is increased by a factor of two. The doubled scanning of the information lowers the noise by the square root of two. The 3-dB increase in the signal/noise ratio that results from this is another positive effect. Users who place more value on speed than on high sensitivity or low noise can double the line rate by illuminating both lines in the sensor simultaneously. The result is a maximum line rate of 140 kHz at resolutions of 2k or 4k, and 70 kHz at a resolution of 8k.
There are no disadvantages for CMOS sensors when presented with strong light intensity, unlike for CCD sensors. These unwanted effects include blooming, which occurs when electrons seep into the neighboring pixel, or smearing, which are light streaks that occur during the shifting of charge carriers. The lack of these artifacts makes CMOS sensors an ideal solution for outdoor applications and indoor applications with poor lighting conditions.
Color Applications: In line scan camera applications where monochrome cameras are insufficient, a variety of color line scan cameras that work with different color systems are available. As with area scan sensors, line scan cameras can gather color data using a Bayer matrix. This is also possible through the use of so-called tri-linear sensors. Both color line scan camera concepts are highly reliable in terms of image quality and are routinely used.
By contrast, the prisms that are inherent to the design of special 3-CCD or 3- CMOS technologies involve a more complicated and expensive process, which makes them rare and only used for absolute special applications. At least two lines must be present for use of the Bayer matrix, with the first line gathering red and green information, and the second line gathering green and blue information.
When using tri-linear sensors, the object‘s color information is gathered using three lines of different colors (RGB). Compared with the dual-line solution with the Bayer matrix, the trilinear solution offers better color resolution, but works at a slower speed due to the higher volume of data (three lines instead of two) at identical resolutions and with an identical interface. Because of the design of the physical camera unit, the two lines in a dual line scan sensor are also positioned right next to one another, with no gap between the lines. This design is not possible for three-cell sensors, since wiring to the pixels must be present to transport the pixel data for the center line.
The gap between the camera can produce fuzzy-looking images in some applications, especially for objects moving at high speeds. Modern cameras have specially developed features to compensate for this, including spatial correction. As a result, this undesirable aspect is no longer an issue for image processing.
CIS Processes for New Developments
Driven by the use of CMOS sensors in cell phone cameras, a stream of new developments related to image noise — for CMOS technology this involves CIS processes — have been launched, leading to enormous leaps in technology. Many of the tricks learned in handling CCD sensors also apply to the newer CMOS technology. The major difference is that implementation moves much more quickly today.
The most important examples for this kind of improvement include, for example, “photogate” and the “pinned photodiode,” whose structure is styled after the CCD pixel. The intelligence available through the CIS process also plays a key role. While CMOS sensors not long ago featured only three transistors per pixel and one rolling shutter, the current standard includes 5-8 transistors per pixel, with a global shutter. The additional transistors do more than just implement the global shutter; they are also used for cutting-edge reading processes, such as correlated double sampling, which provides an additional significant reduction in signal noise.
The rapid improvements to CMOS sensor technology in recent years hints that further advances are likely, particularly in terms of their uses in line scan camera applications. It’s difficult even to make a general statement along the lines of “CMOS sensors are fast, but CCD sensors deliver excellent images.” The latest generation of CMOS sensors shines not just for high speeds and low power consumption, but also for marked improvements in the areas of low noise values, light sensitivity, quantum efficiency, and color systems.
This article was written by Marc Oliver Nehmke, Product Manager, line scan cameras, at Basler AG, Ahrensburg, Germany. For more information, Click Here