CCD imagers were the first products in machine vision and have led the industry for many years. However, CMOS imagers are catching up quickly and are being implemented in many more applications than in previous years. When selecting an imager, a number of issues should be considered, relating both to the different characteristics of CCD/CMOS devices and the requirements of the application.
The technology gap between CCD and CMOS is getting narrower every month and the table below gives a comparison of characteristics of both technologies today.
CCD/CMOS CHARACTERISTICS COMPARISONS SPEED
CMOS has a speed advantage as onboard circuitry allows for low propagation delays and the conversion to digital closer to the pixels. CCD's can achieve high speeds, but at significant cost.
PROGRAMMABLE WINDOWING
CMOS has the ability of dynamically changing X,Y window locations, sizes and frame rates. Some CCD's now allow for several fixed windows that are not programmable.
NON UNIFORMITY
CCD's win this category as CMOS onboard amplifiers improve sensitivity but increase non uniformity, requiring additional software and hardware to reduce non uniformity.
RESPONSE
CMOS imagers have much higher signals in response to the same light levels due to amplifiers being placed at every pixel. This function is expensive and difficult to implement on CCD's.
GLOBAL SHUTTERS
Shuttering is superior in CCD's while maintaining large pixel size. CMOS gives up the larger pixel size for the shutter circuitry. Both families have versions of imagers that improve sensitivity but eliminate the global shutter and generally require the use of a mechanical shutter, which is not practical in most machine vision applications.
SIGNAL TO NOISE
The dynamic range of CCD's is much better due to lower noise. This is a key factor for the use of CCD's, but it should be noted that CMOS is advancing in this area with logarithmic response curves.
BLOOMING
Blooming is an inherent overexposure problem in CCD's that does not exist in CMOS. CCD's generally build in some type of anti-blooming circuitry to reduce the problem. CMOS also allows large dynamic logarithmic response to deal with high-intensity light applications.
INTERFACING
CMOS is a clear winner here with onboard circuitry including A/D and single supply. CCD's generally require multiple bias voltages, clock signals and external A/Ds, making interfacing bulky and expensive.
POWER CONSUMPTION
CCD's consume several times more power then CMOS.
COSTS
CCD technology is a custom process and therefore more expensive than CMOS, which is a standard process used to make most computer chips today.
COMPARISON BY APPLICATION TYPES
As can be seen by the comparison above, CMOS is rapidly catching up to CCD in many areas. However, high quality, lowlight images still give CCD's an advantage, but at higher costs.
We need to segment applications into a few categories in order to be able to decide how CCD and CMOS fit each of these machine vision applications.
STATIONARY OBJECT INSPECTION
These are applications where the object does not move and allows long exposure times or the ambient light is filtered out completely and a high- intensity strobe light source is used to stop or expose the object being inspected. Generally, imagers with global shutters are not required in these applications. Typical applications may be medical applications, automotive body manufacturing lines, license plate reading, some bar/matrix code applications, etc.
CCD — Depending on resolution and sensitivity requirements, the use of consumer type imagers including the large formats that require mechanical shutters, dark conditions or narrow bandwidth filters with strobed light sources. It should be noted that consumer imagers are mainly color and CCD's get very expensive in higher resolution monochrome.
CMOS — Rolling shutter consumer imagers can also be used as above with mechanical or strobed light sources. Again, most consumer imagers are color and hi-resolution monochrome are expensive.
Conclusion — Either imager can work in these applications with CMOS having a lower total cost.
LOW SPEED OBJECT INSPECTION
Typical low-speed 30-60fps applications are label inspection, bottling lines, slower 3D triangulation, bar code/matrix code, etc. Even though the frame rate is low, the shutter time of 1ms is likely needed to do the inspection. A global shutter is generally a requirement here along with a trigger signal to take the image.
CCD — 60fps requires the use of only low-resolution CCD color/monochrome imagers because CCD's generally require the full field of view be captured, limiting the frame rate for high resolution CCD imagers. A trigger signal is needed and multiple objects in FOV requires software extraction of the object of interest.
CMOS — Any resolution CMOS imager that has a global shutter and windowing can be used. The use of a trigger and windowing will allow a single image of the object of interest to be captured, potentially reducing image processing time and increasing frame rate. Free-running applications with no trigger can also be implemented by using the windowed high frame rate modes of CMOS imagers.
Conclusions — CCD has led this application area for many years, but the flexibility of CMOS is challenging now.
HIGH SPEED — 2D SURFACE INSPECTION APPLICATIONS
This category covers color and monochrome surface applications which generally involve one or more synchronized cameras and light sources inspecting various characteristics of a high-speed moving surface. Applications include webs (paper, steel, plastics, glass), roads, wood products, print, etc. These surfaces can be moving at speeds approaching 50m per second and require surface resolutions less then 1 mm. These speeds need a global shutter or exposure times approaching 20 µsec which necessitates intense light sources. Three types of imagers are used in these applications — linear CCD, area CCD and area CMOS. Each has it advantages and disadvantages.
Linear CCD — This is by far the most used imager in this application. Most engineers like to use one expensive large imager to cover the whole FOV to reduce the need to stitch multiple images in software. Generally, this involves mounting the camera with expensive optics far above the surface and lighting the surface with a very intense 100 percent duty cycle light source. Since the imager only sees a narrow row of pixels, most of the light hitting the surface is wasted energy. The large optics required generally need significant distortion correction due to the large FOV. High clock rates and multiple tap imagers are required to access the pixels in many of the high-speed applications.
Area CCD — These imagers can be used in multiple synchronized cameras with stitching of images resulting in much more efficient use of the light with inexpensive optics and short standoff distances. One major advantage to this technique is the use of a low-duty cycle light source as low as 0.2 percent which allows the use of over-driven pulse LED sources mounted at short standoffs. The challenge is manufacturing a large light source to light the full FOV of a CCD because of limited windowing ability. Bayer conversion for color can make the software more complex for stitching, etc. The advantage of low-duty cycle LED light sources now available from UV to IR enables the ability to do multi-spectral imaging in a single surface inspection camera by using different time windows.
Area CMOS — Same characteristics as area CCD with the added benefit that a programmable window enables a smaller low-duty cycle LED light source for a small FOV using a high frame rate window. In addition, CMOS imagers are available at much higher pixel clock speeds equivalent to linear line rates as high as 100K/sec.
Conclusion — There is no question the area imagers offer a clear advantage but require more complicated software and synchronization. The CMOS area imager with programmable windows is the simplest and provides highest speed implementation.
HIGH SPEED — 3D SURFACE INSPECTION APPLICATIONS
This category covers the use of triangulation and involves one or more synchronized monochrome imagers and laser light sources inspecting 3D characteristics of a high-speed, moving surface. Applications include glass, roads, wood products, tire, metal, rubber and plastic extrusions, etc. These 3D measurements are sometimes combined with 2D above and have similar speeds approaching 50m per second. These applications require cameras with effective frame rates approaching 50,000 fps for some applications.
CCD— Currently, most of these applications are not feasible with CCD's. Some of the smaller CCD's are able to read a small fixed window at a few thousand FPS, but clock speeds are limited.
CMOS — The best technology for these applications due to the high clock rate and dynamically programmable windowing ability of CMOS. Very high frame rates can be achieved with standard CMOS imagers with windows that move from frame to frame creating a high-speed moving FOV. Custom imagers have also been produced with multiple taps and onboard 3D processing logic that can achieve these high frame rates.
Conclusion: CMOS is the only technology here.
SUMMARY
In those machine vision applications requiring very high resolution, low light and low noise, use CCD's. Otherwise, a CMOS imager has the flexibility to solve most applications at lower total costs including the highest speed solutions.
Tweet This
9 
