This article is intended to provide some assistance in answering these questions.
Unlike a traditional frame grabber, these units are external peripheral devices that connect externally to the PC on which the actual image processing takes place. Modern PCs are generally already equipped with the requisite external interface. The most widespread interface is the USB interface, but FireWire and Gigabit Ethernet (“GigE”) have also established. These channels opened up new possibilities for getting image data into the main memory of the evaluation PC.
In order to provide support for the various camera technologies available on the market, a large range of different acquisition modules is now available. They are able to capture images from analogue cameras that support common video standards, from CameraLink cameras or from LVDS cameras.
In principle, these boxes operate in the same way as a frame grabber. They capture the camera signal, buffer it, sort it if necessary and then transfer it to the PCs RAM across the interface. Some acquisition modules can be triggered in the same way as a frame grabber, which means that they autonomously cause the camera to capture an image. Some modules also provide pre-processing options such as sorting, Bayer-to-RGB conversion and much more. Some manufacturers even offer boxes which allow several cameras to be connected to a single unit. With these products, the image data is transferred either in parallel or over multiplexers.
The big advantage with acquisition modules is their flexibility. On the one hand, there is a huge variety of mature, readily available analogue, CameraLink and LVDS cameras that can be used, just as with frame grabbers. On the other hand, these devices make it possible to structure image processing systems far more flexibly. Unlike a frame grabber, an acquisition module can be installed in a suitable location up to 100 metres away from the PC, such as in a switching cabinet. This allows the user a far greater degree of freedom during integration. In addition, the evaluation PC can be more compact because a PCI slot is no longer required since an external acquisition module is being used. The boxes themselves are extremely small, having dimensions of approximately 100 x 100 x 40mm. And, of course, external connection to the PC also assists integration.
In addition, acquisition modules also allow multi-camera systems to be set up easily. This is where Gigabit Ethernet really comes into its own, as this technology comes from the world of networking, which means that it is by its very nature, suitable for building a network from virtually any number of sources and PCs. The cables used for Gigabit Ethernet are the same as those already used for networking systems and are therefore already available in industrial quality.
One important disadvantage of using acquisition modules is data reliability. With frame grabbers, the physical connection is established using one or more direct cables. This means that it is always possible to say exactly when and where the signal is being transmitted, how long the data transmission lasts and when the signal arrives.
With GigE, USB and FireWire, on the other hand, the data is split into packets and then sent. Even with an extremely simple topology comprising a single camera, a single acquisition module and a single PC, the precise data transmission path is no longer predictable. If image processing systems are deployed in a network, it is not even possible to guarantee reliable delivery of a data packet within a specific time due to the fact that other devices may have a higher priority.
The second disadvantage associated with acquisition modules is that, unlike CameraLink, analogue or PCI frame grabbers, the actual bandwidth on GigE, USB and FireWire is not necessarily guaranteed: Either the user utilises the entire data rate, in which case transmission is not entirely ensured or they give priority to the reliability of the transmission, which only works at the expense of bandwidth. Although checksums provide a method of identifying data loss, this technology is generally only implemented in the drivers to a limited degree. And to exacerbate the issue, it is not possible with any technology to precisely predict the delay with which the data from the camera will arrive in main memory. Of course, these disadvantages are only of any significance where large volumes of data are involved.
Where time-critical tasks need to be carried out with the assistance of image processing, it is vital to be able to predict the exact timing of the data. And it is here that the traditional frame grabber comes up trumps. No technology offers comparable levels of data reliability coupled with the same flexibility. The frame grabbers from the Canadian manufacturer DALSA, for example, offer the possibility of not only detecting data loss, but also of obtaining precise information about when and where the error occurred during data transfer. This gives the user or the system a chance to react.
In addition to it being a mature technology - tested over many years - a further advantage of frame grabbers when compared with the acquisition modules is the range of functions and efficiency offered by their drivers. With FireWire, USB and Gigabit Ethernet, existing drivers, for instance those provided by Microsoft, may be used, but these are not optimised for image processing. If the manufacturer writes their own drivers, these are not always compatible with the chipsets used in the interface hardware on the PC.
Of course, frame grabbers are not without their disadvantages, and these ought to be considered. They will always demand greater expertise of the user. Among other things, it will always be necessary to open up the PC in order to install the frame grabber, which often causes end users a certain amount of concern. In addition, somewhat larger PCs are required, than for acquisition modules, at least for PCI bus frame grabbers. Since frame grabbers generally deliver a wider range of functions, they are by nature more complex to operate and generally also somewhat more expensive than acquisition modules. Since frame grabbers are always installed in a PC, they also do not provide the flexibility with respect to topology described above.
Cameras in which these technologies have already been directly integrated do not need an acquisition module. However, the user is restricted as to their choice of cameras as such cameras are not yet widely available. In particular, the choice with respect to resolution and refresh rate is still extremely limited. This situation will improve considerably in future; nevertheless, it will undoubtedly be several years before the current range of analogue, CameraLink and LVDS cameras is matched.
Gigabit Ethernet is the major player in this sector. A consortium of several manufacturers of frame grabbers, cameras and software worked on a generally applicable image processing standard based on Gigabit Ethernet. This standard means that this protocol, which features data rates up to 1000 Mbyte/s and which is inherently network-capable, has the greatest likelihood of widespread use for image processing.
The following companies belong to this consortium: Adimec, Atmel, Basler, CyberOptics, JAI, Matrox, National Instruments, PhotonFocus, Teledyne DALSA and STEMMER IMAGING. The objective of the group is to present a working prototype as soon as possible. More information on the GigE consortium: www.machinevisiononline.org).
Although handling the components has become much easier, it still remains important for users to be advised by a competent specialist as to the precise technology that will best solve the user's task.