The Automated Imaging Association (AIA) just released GigE Vision 2.0 in December 2011. This is the fourth version of the popular camera interface standard based on Ethernet. This new release tries to build on previous successes by improving the connection to faster sensors. This is performed in a number of ways.
FasterThe new version formally introduces support for 10 Gigabit Ethernet. Even though it is possible to support 10GigE with previous versions of the standard, since GigE Vision is built on top of Ethernet, the original text did not explicitly state this is allowed. This has been changed. Using 10GigE is a brute force approach to increase the transfer throughput by augmenting the clock speed on the cable. This can be done for your typical copper cable (CAT-6a) or using fiber optic. A different approach is to combine more than one cable but keep the clock rate the same; this is called link aggregation. Hence, by using 2 gigabit/sec cables, a GigE Vision camera can double its maximum throughput from 125 to 250 MB/s. This type of camera has been available for some time, but now GigE Vision 2.0 clearly defines the way to configure it to facilitate interoperability. The power of the GigE Vision logo is interoperability to avoid bad surprises during system integration.
Going to higher transfer speed might create a bottleneck between a transmitter and a receiver. To better cope with the large amount of data, a receiver can use the PAUSE mechanism to notify the sender that its “receive buffers” are full. This avoids sending data that cannot be accepted by the receiver and eliminates packet resend requests that would themselves contribute to augment the bandwidth bottleneck.
SmallerFrom the start, the main objective of GigE Vision 2.0 was to increase the amount of information that can be transferred. Augmenting the total physical bandwidth, as shown earlier, is one approach. A different approach is to reduce the amount of data to transfer. To this end, GigE Vision 2.0 introduces a number of data compression schemes: JPEG, JPEG 2000 and H.264. Even though image compression might not be lossless (i.e. some amount of the information is lost due to the compression step), these compression standards have proven over time to perform fairly well and might be an interesting alternative to improve transmission bandwidth while keeping system cost low. Some applications require archiving the images. Native data compression within the transmission standard eliminates additional burden to implement the data compression.
Transfer overhead is higher for small images since GigE Vision uses some packets to announce the format of the next image. Taking into consideration that Ethernet allows for jumbo packets, GigE Vision 2.0 introduces a so-called “All-in Transmission” mode where a full image can be transmitted using a single packet. By regrouping all the image information into a single packet when it fits, a lot of the packetization overheads are eliminated, helping streamline transfer. This is expected to be useful when small regions of interest need to be transferred at a high rate from CMOS sensors. This technique also could be used successfully for some line scan applications.
And MoreGigE Vision 2.0 also looks at improving the real-time synchronization of multi-camera systems. It leverages IEEE 1588 Precision Time Protocol to allow each camera on the network to be synchronized up to 1 µs. This way, it is now possible to schedule actions to be executed simultaneously by the various devices attached to the network. And this 64-bit timestamp is attached to the image to help applications correlate data coming from multiple cameras.
Sensor vendors are using multiple taps in order to increase the readout throughput. In previous versions of GigE Vision, the camera had the burden of reconstructing the image locally in raster-scan format before transmission. GigE Vision 2.0 introduces a format called “multi-zone Image” to simplify this operation. An image can be divided into horizontal bands and, even though each band must be transmitted in raster-scan, the camera is now allowed to transmit packets from different bands in any order.
GigE Vision devices can be much more than a simple camera. The standard supports various payload formats (uncompressed image, compressed image, file transfer) and, since version 1.2, can even be used by non-streamable devices, such as I/O boxes that provide a GenICam™ interface to read and write registers. This speeds up development since the same function calls can be used to configure any GigE Vision device, and not only cameras. Continuing in this trend, GigE Vision improves on the device discovery process by allowing usage of multicast DNS (mDNS). Previous versions of GigE Vision only allowed for a polling mechanism to enumerate devices. mDNS enables a more efficient approach where devices can announce their presence to the network. This makes GigE Vision 2.0 more network- friendly.
Pixels EverywhereIn recent years, many new machine vision camera standards started to emerge: CoaXPress, Camera Link HS and USB3 Vision, to name a few. It becomes important to ensure a certain level of coherency between these standards in order to facilitate their co-existence in machine vision systems. To help achieve this goal, the GigE Vision committee decided to generalize the definition of its pixel formats and create a separate document called “Pixel Format Naming Convention.” This document, maintained by the AIA, provides the layout of the various supported pixel formats and enables re-usability across the camera interface standards.
GigE Vision 2.0 also introduces a number of new pixel formats, including 1-bit, 2-bit and 4-bit monochrome and various YCbCr formats derived from ITU BT.601 and ITU BT.709.
ConclusionGigE Vision 2.0 has arrived. It will enable the machine vision industry to expand beyond its traditional markets by offering functionality not available in a typical machine vision camera. GigE Vision retains its long cable and option for power over the cable. By improving the transfer speed, release 2.0 enables supporting the faster sensors found today.V&S
Tech TipsBy using 2 gigabit/sec cables, a GigE Vision camera can double its maximum throughput from 125 to 250 MB/s.
GigE Vision 2.0 introduces a number of data compression schemes: JPEG, JPEG 2000 and H.264.
It is now possible to schedule actions to be executed simultaneously by the various devices attached to the network.
A 64-bit timestamp is attached to the image to help applications correlate data coming from multiple cameras.