Machine Vision News
Vol. 10, 2005
Camera Link Reaches Full Potential with PCI Express
Four years ago, camera makers and image acquisition hardware providers standardized the cabling and communication between cameras and frame grabbers with Camera Link. The full configuration of Camera Link supports eight taps, or input channels, and can acquire images up to 680 MB/s.
Understandably, at this speed the PCI bus becomes a bottle neck, transferring data at a maximum of 130 MB/s. While this maximum speed is no where near Full Configuration Camera Link, it has been enough to solve many machine vision application challenges. However, over the last few years, faster cameras and faster processors have highlighted the need for a faster bus between the frame grabber and the PC. Now, with the introduction of PCI Express and the inclusion of the standard in every Intel chipset, there exists a widely available, low-cost bus for acquiring images at high speeds with PCI Express frame grabbers, such as the NI PCIe-1429.
The NI PCIe-1429 is the industryís first PCI Express image acquisition board for high-throughput vision applications. With the NI PCIe-1429 image acquisition board, you now can acquire images at the highest speeds, resolutions and bit depths available for Camera Link cameras to perform demanding imaging applications such as synchronized data and image acquisition, fault analysis and advanced motion tracking.
In the past, acquiring images at high-speeds required devices with banks of expensive onboard memory that could only acquire images for short periods of time, or specialized buses such as the PCI 64/66 or PCI-X, which are not supported by standard PC chipsets. With Intelís standardization on PCI Express and the NI PCIe-1429 image acquisition board, you now can acquire high-speed data indefinitely through a standard PC bus.
Camera Link is an industrial high-speed serial data and cabling standard developed by National Instruments, camera vendors, and other image acquisition companies. Created for easy connectivity between the PC and the camera, Camera Link provides simple, flexible cabling for high-speed, high-resolution digital cameras. A Camera Link cable is a slender 26-pin cable with 24-bit data, clock, and enables as well as control signals. You can control camera functionality by asynchronous serial control or LVDS differential lines through a Camera Link cable.
Camera Link comes in three configurations Ė base, medium, and full. The base configuration uses three eight-bit taps, or input channels, to acquire up to 24 bits of data at a rate of 85 MHz or 255 MB/s. The full configuration supports eight taps and can acquire images up to 680 MB/s.
Integrating a vision application into the rest of a test or automation system often requires more than just one trigger line. However, with full-configuration Camera Link requiring two large connectors, there isnít room on the boardís faceplate to house more than one digital I/O line. To address this challenge, additional digital I/O lines for advanced triggering, pulse-train outputs and isolated DIO also are available with the IMAQ 1000 expansion board. With the boardís four-lane PCI Express configuration, engineers and scientists can acquire at the full Camera Link bandwidth of 680 MB/s. In addition, they can synchronize other data acquisition measurements with each acquired image to analyze activities frame by frame in data-intensive applications such as crash tests.
Images acquired at 1.000 frames per second with a Basler A504 high speed camera
The NI PCIe-1429 image acquisition board is ideal for many industrial, life science and biomedical imaging applications. For instance, engineers and scientists can use the board to perform fault analysis by setting up a stop trigger to record images before and after an event on the factory floor. Also, they can use high-speed imaging to perform particle image velocimetry, track the movement intricacies in gait analysis or measure the stimulus responses in heart valves or eye corneas.
About PCI Express
Think of the new PCI Express architecture as a serialized version of PCI. From a software perspective, registers, DMAs, and interrupts are basically unchanged. This facilitates a smooth migration from PCI to PCI Express. PCI Express offers up to 32 serialized lane pairs running at 2.5 GHz. The new NI PCIe-1429 image acquisition board uses a four-lane PCI Express configuration, which allows for a bandwidth of 800 MB/s. It is no surprise that bandwidth hungry vision applications, like motion tracking and particle velocimetry, have already benefited from PCI Express. However, with the impending PC dominance of PCI Express, it is reasonable to assume that in the near future most PC-based image acquisition will use the PCI Express bus and that the PCIe-1429 is just the first of many PCI Express-based imaging devices.
Machine Vision for the Rest of Us
Donít get bought into the idea that you need a machine vision specialist or integrator to build your visual inspection system. With new technologies aimed at simplifying machine vision development, anyone can design, build, and deploy seemingly sophisticated machine vision applications without needing a degree in image processing.
You shouldnít have to be a computer scientist to program a machine vision application. If youíre not experienced in a programming language like C++ or LabVIEW, then you should consider using software that lets you configure your machine vision application without programming. These types of configuration software usually consist of friendly, menu-driven interfaces that enable you to acquire images from your camera, explore different machine vision functions like pattern matching and gauging, and experiment with what-if conditions without ever having to program. Some configuration software packages go beyond simple applications by letting you set up pass/fail conditions on each inspection and route this information to relays, lighting, and databases. National Instrumentsí Vision Builder for Automated Inspection is a configurable software tool that even lets you download your application to a real-time operating system for more robust and time-critical applications.
While configuring a machine vision application is fast and easy, there are drawbacks. To highly customize your algorithms, build a custom user interface, or integrate with motion control or data acquisition system, a programming language, like LabVIEW is often necessary. Vision Builder for Automated Inspection generates working LabVIEW code with a single mouse click. So in the future, if you find you need to upgrade your software beyond the capabilities of your configuration environment, you can switch to LabVIEW without having to rewrite you inspection.
IEEE 1394 Cameras:
As new cameras become available, you should be able to take advantage of new camera features without having to rewrite your application, buy new cables, or reinvest in new image acquisition hardware. IEEE 1394 (FireWire) cameras make this possible while making machine vision simpler and cheaper. Rather than purchase a specialized image acquisition hardware device for your PC (often called a frame grabber), you can use an inexpensive, consumer-grade IEEE 1394 controller card, available at most PC electronic stores. The cabling for FireWire cameras is very simple; all you need is the standard, 6-pin cable between the controller card and camera which provides lines for both power and for data. The 400 Mb/s bandwidth of IEEE 1394 can handle all but the most demanding machine vision applications, and with more than 60 available industrial FireWire cameras from makers such as Sony and Basler, there is no shortage of selection. Since the cabling and interface is standard for IEEE 1394 cameras (through a spec called DCAM), there is no hassle associated with creating custom cables or learning vendor specific configuration software. Also, IEEE 1394 cameras are hot swappable, so you donít even need to shut down your computer to upgrade cameras!
Intelligent Vision Systems:
So far weíve talked about how to simplify your application software by using configuration environments instead of programming languages and how to reduce the cost and complexity of your image acquisition with standardized IEEE-1394 cameras. The final piece of a typical machine vision system is the processing unit. Most often this is a PC, which allows you to take advantage of ever-increasing processor performance at a relatively low cost. However, PCs are not always appropriate for applications in industrial environments, tight spaces, or remote locations where you need a system what can withstand extended temperatures, fit into small spaces, or reliably work over long periods of time. Smart cameras and compact vision systems are designed to address these conditions. Smart cameras are machine vision devices that combine an imaging sensor with a processing unit. Compact vision systems consist of a robust image processing unit with a tethered camera attached. Smart cameras tend to be slightly smaller than compact vision systems, but they usually arenít as powerful or flexible. For example, applications that require more than one camera are more expensive and complicated with smart cameras because you must purchase a smart camera for each camera view, where with many compact vision systems, the user only needs one processing unit for several cameras. An example of this is the National Instruments CVS-1454, Compact Vision System, which is a powerful and rugged vision system that can connect to multiple IEEE-1394 cameras.
One final thing to keep in mind when exploring machine vision options is technical support and application assistance. While new tools like configurable software and IEEE-1394 cameras take away much of the complexity in building your own machine vision system, it is often very helpful to have a team of experts available to assist you with technical issues.
The machine vision industry is trending toward configurable software, standardized cameras, and specialized hardware. All of these new technologies are enabling end-users around the world to build custom machine vision application both quickly and cheaply without needing special expertise in machine vision applications. To find our more about developing your own machine vision application, just go to National Instruments machine vision website at www.ni.com/vision.