Vision standard clearly supports 10 GigE
19 April 2011
Suzanne Gill reports on some of the main features expected from the next version of the GigE Vision standard, which will be introduced later this year.
The introduction of Gigabit Ethernet (GigE) changed the way cameras are used in machine vision applications and its acceptance and use in industrial imaging applications continues to grow.
The speed of Gigabit Ethernet means that it is now possible to transfer image data to PCs quickly to make that ‘real-time’ application possible using GigE Vision based cameras, where traditionally only dedicated links, such as CameraLink, have been available.
The GigE Vision standard was created to provide a version of GigE that is ‘tuned’ to the requirements of the machine vision industry. The standard was first released in 2006 by the Automation Imaging Association (AIA) and has become one of the dominant camera standards in machine vision today, exceeding CameraLink in the number of units shipped in 2009.
The standard has been through several iterations since its launch, the current version, 1.2, was released in January 2010. This version allowed non-streamable devices, such as I/O boxes to be controlled using he same software interface, allowing new classes of GigE Vision devices to be controlled, in addition to cameras, easing integration of a variety of GigE Vision components in the machine vision system.
With the launch of GigE Vision 2.0, later in 2011, comes the promise of improvements in speed transfer, offering further opportunities for real-time machine vision systems.
Eric Carey, R&D director for DALSA Montreal and chairman for the GigE Vision Committee at AIA, explains more about the main focus for 2.0. He said: “The mandate given to the Gig E Technical Committee by the AIA was to focus on increasing transfer speed. We looked into five main features to help achieve this. The first being 10 Gigabit Ethernet. Even though it is possible to support this using version 1.2 or even 1.0 with Version 2.0 we wanted to be able to say, explicitly, that 10 Gigabit Ethernet will be supported.
“Gig E Vision is, essentially, a pair of protocols that reside on UDP/IP which are, basically, Ethernet based protocols. So, any actual physical interface that can support Ethernet can be used with GigE Vision too. There will also be a fibre-optic version of GigE that will include Power over Ethernet and a bunch of other Ethernet based standards. We want to make it very clear that 10 Gigabit Ethernet is, indeed, supported!”
Version 2.0 will also introduce ‘link aggregation’ which enables multiple Ethernet cables to share traffic. Carey explains: “By putting two cables on the camera you can basically double your bandwidth. There are already some products available that support link aggregation but we want to make sure that the implementation by camera vendors is interoperation and that there are no link issues when users try to interface link aggregated devices to the GigE Vision software.”
Another feature that will be tackled with Version 2.0 is image compression. Carey said: “The aim is not to invent a new image compression scheme, but essentially to provide a pipe into which compressed data can be placed. We are going to support three basic image compression schemes. Jpeg, Jpeg 2000, and H.264, all well known standards.
“We are also looking to simplify situations where the image size is small and it can fit into a single packet. With GigE Vision 1.2 when you transmit an image it takes a minimum of three packets. The first is the data leader that gives information about the image that is coming. At lease one packet is needed for the pixels and finally you need the data trailer – a packet that says the image is completed. So, for very small images, or when you are using ‘area of interest’ GigE Vision 2.0 will enable all the information to be put into a single packet.
“Faster speeds require a better handshake between the transceiver and the transmitter to ensure that packets are not being lost or that you are not allowing too many packets are coming too fast at the receiver. This will be managed simply by reusing the IEEE pause mechanism that is an option in Ethernet.“
Other new functionality is expected to include improved support for a multi-type sensors – allowing the transmission of out of order packets. This will give users more flexibility in the order that pixel packets are sent over the network. Version 2.0 will also introduce support for IP V6, the next version of the Internet protocol, and there will be a standard pixel naming format to enable pixel format to be reused by other camera interfaces.
Latency and jitter issues
Real-time applications will always have to deal with the issues of latency and jitter and GigE Vision is no different.
Latency is the time taken between the start and completion of a task, and jitter is the time variation when the same task is executed multiple times.
When creating a real-time system it is important to define the latency and jitter requirements and ensure that the system can operate reliably and deterministically within those specifications. Carey explains: “There is always latency because the camera needs to expose the object. It needs to send the image to the PC, either over Ethernet or via a frame grabber. If your frame rate is 30 frames per second the readout time will be around 33ms. There is a direct relationship between the readout time and the latency to transmit the image. This latency comes from the analogue to digital converter used to convert the image into digital values –pixels. Once the PC has received the image it needs to process it and this is performed using image processing algorithms which can take anything between 1ms and 100ms to come up with a result.
“It is true to say that GigE Vision introduces more jitter than other systems and that is because the frame grabber has been removed from the equation, instead relying more on the PC to perform the acquisition. However, in most applications GigE Vision is good enough. Even with an additional 1 -1.3ms of jitter, the latency is pretty much identical and is mostly dictated by the camera readout time. Overall this jitter is very short and can be managed within most image vision systems. Overall, GigE Vision can offer a good solution for many applications, with large cost benefits – a GigE camera does not cost much more than a CameraLink camera for the same bandwidth and you do need a frame grabber.
Another important benefit of GigE solutions is cable length. With CameraLink you are limited in terms of cable length – normally between 7 and 10m. With GigE, because of Ethernet, it is possible to go up to 100m, which can make a big difference in many vision systems, where the camera is not that close to the PC.
Full speed ahead
It is hoped that the formal introduction of 10 Gig E as part of the standard will accelerate adoption of 10GigE.
Pleora Technologies, a co-founder of the GigE Vision standard, tells CEE that a number of leading manufacturers and system integrators are already developing with 10 GigE in mind. The company is already working with customers to integrate its soon-to-be-released 10 GigE video interface in systems and industrial cameras.
Vincent Rowley, System Architect at Pleora, said: “The leading-edge nature of the systems that our customers design means that we are already seeing demand for the speed of 10 GigE interfaces because they allow high-resolution without sacrificing frame rate in applications such as precise quality inspection or medical-related analysis.
Of the high-speed video interfaces available, 10 GigE is the only one to enable high-value, next-generation innovations, such as distributed inspection systems, made possible through networked video. This capacity lengthens the life span of a system, and therefore increases the end-user’s return on investment.
“Networking video cuts down on overall system cost and complexity because it reduces the hardware required per camera,” explains Rowley. “With 10 GigE, cabling can be simplified since more than one high-speed camera can share the same Ethernet link.
“However, perhaps more importantly, is the fact that more design and implementation options are made available to system integrators. Networked video over 10 GigE enables designers o take advantage of different topologies, including traditional point-to-point, point-to-multipoint, and multipoint-to-multipoint video connectivity options.”
In Figure 1, a plasma screen inspection system is designed with GigE Vision over 10 GigE. By distributing video over a 10 GigE network, each camera is able to send images to each of a bank of PCs, with each PC searching for a specific defect. This increases inspection speed and provides exceptional value to manufacturers seeking to increase competitive advantage. Additionally, the introduction of new inspection algorithms is simplified. In this scenario, a second bank of PCs could be trained by providing real-time image data from the cameras, all while not interrupting the in-service system.
Most Viewed Articles...