The main features of GigE V ision have been covered previously, but as a quick recap some of the points of note are:
Long cable lengths – upto 90m between links with the capability to use fibre - optics to go much further between links , all this with standard consumer components.
Network topologies – one camera to one PC , many came ras to one PC and one camera to many PCs – ethernet - based transmissio ns has opened possiblities that were not feasible with framegrabber - based systems and gone much further than was possible with USB and Firewire - based systems.
Self-describing cameras – Genicam means that a de vice must describe itself to a connecting PC by means of an xml file. This replaces the need for conf iguration or camera files, for example for describing the format that data streams fr o m a camera. P ower over Ethernet PoE (Power o ver Ethernet) is a standard consumer technology that allows upto 15.4W of DC power without sacrificing the cable - length capabilities of GigE . PoE - enabled G igE camera include IDS uEye CP cameras and optionally Stemmer Imaging ’ s CVC GE cameras and the AVT Mant a and GT ranges . The advantage is that with a PoE - enabled NIC it is only ne cessary to run a single cable to the camera.
Since the advent of Gigabit ethernet cameras, the general advice is to use Intel Pro cards as these offer the most performance options. However, Intel - branded cards do not have PoE capabilities , but Intel chipsets are available in third - party cards which do implement PoE, such as ADLINK ’ s GIE62+ and GIE 64+ cards (two and four channel, re s pe c tively) .
The possibilities offered by GigE for large numbers of cameras per PC c an quickly make processing power the bottleneck . For colour (Bayer) cameras that are close to th e GigE bandwidth limit , such as DALSA’s high- speed Genie HC range, even Bayer processing can lead to significant processing overheads. In light of this Silicon Software has a range of GigE Vision framegrabbers to reduce the burden. At the lower end of the scale are the A-series boards, these are delivered with va rious on - board capabilites such as Bayer conversion, look - up tables to improve contrast, white balancing, digital I/O and filtering – all with minimal CPU interaction. These boards have 4 channels and are available with PoE and can transfer upto 760MB/s (remember that Bayer conversion multiplies the amount of data from a camera ) .
For higher processing requirements, Silicon Software has it’s V- series boards. These have a user-programmable FPGA (supporting SiliconSoftware’s Smart Applets and Visual Applets programming interfaces) as well as twice as much on-board memory as the A-series boards. These are designed to massively offload the CPU from vision proc essing tasks, for example the creation of 3D data from laser sheet-of - light systems or processing that must be carried out in a deterministic time such as particularly high - speed or high - throughput systems. If the standard board’s FPGA is not sufficient it isi also possible to add modules to extend the hardware resources of the board. Once again digital I/O is built into the board, allowing it to work directly with the FPGA design if requ ired. GigEVision 2.0 At the time of writing, th e r elease of G igE Vision 2.0 is imminent. This brings further feature -support to GigE Vision, including
Firewall traversal, where a small amount of data is returned to the camera to show that the connection to the camera is a two - way connection, rather than a malicious attack.
IEEE 1588 Precision Time Protocol, where each device uses a synchronisedclock (synchronised in terms of both time and frequency).
Compressed image streaming support, including J PEG, JPEG2000 and H264 for mats.
Multi-zone images, for example for CMOS sensors that can output data from multiple ROIs.