Many users of CCD sensors are asking about the advantages of the latest CMOS sensors, particularly if they have been using CCD-based cameras. The two sensor technologies, a comparison of the new CMOS sensors and existing CCD sensors, and tips for when it makes sense to select a new camera with CMOS sensors are provided in this article, as well as what to expect after integration.

What is the Difference?

A CMOS sensor.
Charge coupled device (CCD) and complementary metal oxide semiconductor (CMOS) image sensors are currently on the market. Their task is to transform light (photons) into electrical signals (electrons). This information is transmitted by both sensor types using different methods, and the design of each is also fundamentally different.

In CCD sensors, the charges of the light-sensitive pixels are shifted and converted into signals. The charges of the pixels, which are created by exposure to a semiconductor, are transported to a central A/D converter with the support of many very small shifting operations (vertical and horizontal shift registers), similar to that of a “bucket chain.” The transfer of the charges is forced with the support of electrical fields, which are created by electrodes in the sensors.

Figure 1. Design of a CCD sensor (left) and a CMOS sensor (right). The charge is shifted further pixel-by-pixel in the CCD sensor. In the CMOS sensor, the charge of each pixel is directly converted to a voltage and read, which makes the CMOS sensor noticeably faster.
In the CMOS sensors, a capacitor as a charge storage is put in parallel to each individual pixel. This capacitor is charged with the exposure of each pixel by its photoelectric current. The voltage created in the capacitor is proportional to the brightness and the exposure time. In contrast to CCDs, the electrons captured in the capacitors by the exposure of the sensor to light are not shifted to a single output amplifier, but are transformed into a measurable voltage directly at the source by means of each pixel’s own associated electronic circuit. This voltage can then be made available to the analog signal processor.

Figure 2. A 4-tap sensor
By using additional electronic circuits per pixel, each pixel can be addressed without the charge having to be shifted, as with CCDs. This results in the image information being able to be read much more quickly than with CCD sensors, and artifacts due to overexposure such as blooming and smearing occur far less frequently or not at all. The disadvantage is that the additional space required for each pixel’s electronic circuit is not provided as a light-sensitive area. The portion of the light-sensitive area on the sensor surface (defined by the fill factor) is then smaller than that of the CCD sensor. Theoretically, for this reason, fewer photons for the image information can be collected. There are methods, however, for lessening this disadvantage.

The Multi-Tap CCD Sensor

The transfer of the charge in the CCD sensor requires a great deal of time. This is particularly a disadvantage with high-resolution sensors in which the charging must be fed into the central amplifier by many shifting operations based on the large number of pixels. This narrowly limits the maximum frame rate. The technical response to this problem is the multi-tap sensor.

Table 1. CCD Vs. CMOS Sensors
In the multi-tap sensor, the sensor surface is divided into multiple tap areas. Each tap area has its own electronic circuit (the tap) for creating a signal, and an individual output for each of the tap areas. The image information from the tap areas is shifted, amplified, and selected by the taps simultaneously over shorter distances, and is therefore faster. These areas must later be reassembled into an image. The multi-tap process provides high resolution and speed, but also has the disadvantage that it is very complex. The individual tap electronic circuits must carefully be adjusted on top of one another. Even the smallest deviations result in visible differences in the image, which are visible to the human eye because of the distinct boundaries of the tap areas. The energy consumption of multi-tap sensors is generally greater, which leads to increased heat generation. This has a tendency to increase the noise of the sensors, especially as appropriate to make cooling measures necessary. Due to its internal design, high speed and high resolution can be realized with CMOS sensors without the necessity for using multi-tap architecture.

Why the Latest CMOS Sensors are Superior

Only very recently have high-resolution global shutter CMOS sensors been available. Many sensors previously were based only on the rolling shutter. The image quality of many CMOS sensors today is also superior to the image quality of CCD sensors. This is also one of the reasons why a world market leader of CCD sensors has discontinued them and is concentrating entirely on CMOS in the future.

Figure 3. The image on top was taken with a 4-tap CCD sensor; the bottom image was taken with the new CMOS sensor. In the bottom image, a clearly higher dynamic range can be seen very well as it allows for the driver, as well as the license plate, to be better recognizable in the same picture. Additionally, with nearly the same settings, the sensor is clearly more sensitive and there are correspondingly more details in the background.
In the example in Figure 3, the image on top was taken with a 4-tap CCD sensor. The bottom image was taken with a new CMOS sensor. In the bottom image, a clearly higher dynamic range can be seen, allowing for better recognition of the driver and the license plate in the same picture. Additionally, with nearly the same settings, the sensor is clearly more sensitive, and there is correspondingly more detail in the background.

A good comparison of image quality is the tendency for increased noise of one sensor over another with the same settings. The signal-to-noise ratio (SNR) is best determined by use of the image gray level spectrum (Figure 4 top) of a homogeneous light gray surface (Figure 4 bottom). The lower the width of the spectrum of the gray values, the better. In the figure, the images of the light gray surface and their gray level spectra also are shown – on the left for the CCD Global Shutter Sensor from Sony, and on the right for the CMOS Global Shutter Sensor from e2V. The influence of the pixel size has been eliminated.

When to Consider Changing Camera Technology

Figure 4. Comparison of image quality of CCD and CMOS sensors.
If one or more of the following questions can be answered with “yes,” then it is time to change to CMOS technology. This applies equally to existing systems and new systems to be developed:

  • Would I like to achieve an increase in the performance in my system by higher frame rates?
  • Would I like to achieve an increase in the performance to also be able to see more under difficult light conditions?
  • Does the heat generation in the camera present a problem? Does it have to be excessively cooled?
  • Are image artifacts such as visible lines, blooming, or smearing a problem?

What About Integration?

If the decision has been made to change the sensor technology, a couple of things should be kept in mind to ensure a quick and effective integration. The complete and highly complex sensor integration, including the optimization of the image quality, is part of the core know-how of the manufacturer, and already is established by the time you have the CMOS camera in your hands. In selecting a camera, the user only has to be concerned with the “exterior” points.

Determine the right camera for the necessary resolution, sensor, and pixel size. In practice, resolution describes a measurement of how large the smallest possible distance between two lines or points may be so that they can still be perceived as separate from one another within the image. So what is meant when you read a data sheet and it states 2048 × 1088? This information refers to the number of image points (pixels) per line; in this case, 2048 pixels for the horizontal lines and 1088 pixels in the vertical lines of the image. Multiplied with one another the result is a resolution of 2,228,224 pixels, or 2.2 megapixels (million pixels, MP). A simple formula is used to determine which resolution is required for your application (See Figure 5).

Figure 5.
The required resolution depends on which details you want recognized in the image. Large surfaces, both on the sensor and in the individual pixels, offer more space to capture light. Light is the signal from which the sensor generates and processes the image information. The more surface area available, the better the SNR, in particular for large pixels with 3.5 mm. The better the SNR, the higher the image quality. A good value is within the range of 42 dB.

Another benefit of a large sensor is the larger space on which pixels can fit, which produces a higher resolution. The actual advantage here is that the individual pixels are still always large enough to guarantee a good SNR, as opposed to smaller sensors on which less surface is available for smaller pixels.

Keep in mind that large sensors and many large pixels without the corresponding lens is only half the story. They can only achieve their full potential when combined with a suitable lens that is capable of depicting such high levels of resolution. Large sensors are always more costly, since more space always means more silicon.

Define the required camera interfaces. This decision depends on, among other things, the required cable lengths, bandwidth, speed, and real-time requirements, and the availability of the PC hardware.

Table 2.The various interfaces in overview.
Here it is important to note that many CMOS sensors provide a high data rate and thus require a large bandwidth. For this reason, a camera interface should be selected that supports a high bandwidth and has a cost-effective infrastructure (e. g. GigE or USB 3.0). Then the system is well set up for the future if higher frame rates should once again increase the performance.

Table 2 provides a concise graphic overview of the current camera interfaces, and their advantages and disadvantages. GigE Vision and USB 3.0 will dominate the interface market for some time, so a change to CMOS is the best choice.

Selection of lens and lighting. If one decides on a new sensor format, then a new lens should come with it. The lighting must also be adapted if the new sensor has a different sensitivity. In many cases, it is possible to increase performance and also reduce costs. Smaller pixel sizes also allow smaller lens formats that are available for a more favorable price (as long as the optical solution also fits). An example is the 1/2” lenses that provide more than 5 MP resolution.

Integration expenses for software and camera control. Cameras that conform to a current standard such as GenICam, or interface standards such as USB3 Vision or GigE Vision, are generally easy to integrate. Previous programming can possibly be maintained, and only the necessary recording parameters are adapted. If the previous solution did not correspond to a standard, the integration is bound to be somewhat costlier, but it can still be worthwhile. The new solution should be prepared enough for the future so that other less expensive cameras can also be integrated at any time.

Selection of the next suitable camera. With this checklist, you can determine the right CMOS camera. In doing so, keep in mind the following points with respect to your old CCD solution:

  • Optical format and pixel size: It should stay the same if no change in the lens is desired. It can be smaller if the sensitivity is greater and a lens change is possible.
  • Frame rate: Ideally it should be higher (to achieve performance improvements in the system).
  • EMVA data: Should be the same or better.
  • Sensitivity/wavelength: It should be similar if the lighting cannot be adapted.
  • Design size of the camera: Should be the same or smaller.
  • In-camera firmware functions: These should be compared in detail if particular firmware functions have been used up to now. Modern CMOS-based cameras usually offer more functions. Examples include sharpening or noise reduction algorithms.
  • Software and programming: If the processing had previously been done with software that conformed to standards (e. g. GenICam and GigE Vision), then the same compatibility should be used so that minimal adaptation is made to the programming. If proprietary software was used, more time should be allowed for making adaptations to the programming. Changing to software that conforms to standards is therefore recommended.
  • Camera interface: The same interface should be used if USB 3.0 or GigE; for older interfaces or grabber-based interfaces, making a change should be considered to reduce system costs and/or to have a more sustainable design for the future.

Conclusion

Modern CMOS sensors are generally superior to multi-tap CCD or standard CCD sensors. And that is not only with regard to the price, but also because of unambiguous technical advantages such as higher speeds, higher resolutions, fewer picture interferences, or negligible heat generation. The integration of new CMOS-based cameras as a replacement or alternative to CCD sensors can also be a simple process, especially if the user selects hardware and software that conforms to standards. This means that, for example, the throughput for inspected parts may clearly increase for a less expensive camera and inspection system without having to simultaneously make cuts in the image quality.

This article was written by Rene von Fintel, Head of Product Management Mainstream for Basler AG, Ahrensburg, Germany. For more information, Click Here .

NASA Tech Briefs Magazine

This article first appeared in the October, 2015 issue of NASA Tech Briefs Magazine.

Read more articles from this issue here.

Read more articles from the archives here.