The most simple setup is a single camera on a dedicated NIC (no other network traffic) with a point to point network connection. In order to retrieve their IP address, GEV cameras are configured to use DHCP by default and the same applies when configuring a NIC under Microsoft Windows. If there is no DHCP server (which is likely on a point to point connection) the camera and the NIC will, after a while, fall back to a mode called LLA. In this mode both ends (camera and host) will pick a random IP address.
If you want to eliminate the time delay until they fall back to LLA you have to assign static IP addresses.
In this type of scenario the bandwidth and data latency is optimal. Depending on which components are used and your system specification you can expect to get over 100MB/sec.
With a camera connected to a corporate network the scenario is only slightly different since a corporate network will usually have a DHCP server and both the camera and the PC will get their IP address from it.
On the other hand, you cannot be sure that there won’t be other cameras on the network. In large set-ups there is a good chance that the system will address the wrong device.
The bandwidth usage on such a system might be critical since it will affect the corporate network.
In this setup the IP address assignment is similar to the previous sections. The main concern in this case is the bandwidth.
If in such a system the peak bandwidth is above one Gbit you have to make sure that either the switches can buffer enough data to sustain that peak or you have to limit the bandwidth of the cameras by setting the InterPacket delay of the cameras.
The InterPacket delay is a mandatory bootstrap register of a GEV device. It can be set using the CV GenApiGrid Control (NodeMap) and then saving this to the profile of the camera to restore the setting on the next boot up. For some cameras this is controlled over the StreamBytesPerSecond parameter.
With multiple network interfaces you get better performance than with a single NIC but the IP-address assignment is a little bit more complicated. You have to make sure that each NIC is running on a separate subnet.
If you have two NIC's running on the same subnet under windows, one cannot predict which NIC will be used to send the packet. If the packet is sent via the “wrong” NIC which is not connected to the correct camera, this camera will never see the packet and a connection will not be established.