[Documentation] [TitleIndex] [WordIndex

Roboception GenICam Convenience Layer

This package combines the Roboception convenience layer for images with the GenICam reference implementation and a GigE Vision transport layer. It is a self contained package that permits configuration and image streaming of GenICam / GigE Vision 2.0 compatible cameras like the Roboception rc_visard. The API is based on C++ 11 and can be compiled under Linux and Windows.

This package also provides some tools that can be called from the command line for discovering cameras, changing their configuration and streaming images. Although the tools are meant to be useful when working in a shell or in a script, their main purpose is to serve as example on how to use the API for reading and setting parameters, streaming and synchronizing images.


The tools do not offer a graphical user interface. They are meant to be called from a shell (e.g. Power Shell under Windows) or script and controlled by command line parameters. Calling the tools without any parameters prints a help text on the standard output.

NOTE: If any tool returns the error 'No transport layers found in path ...', then read the section 'Transport Layer' below.

Device ID

There are multiple ways of specifying an ID to identify a device.

1. In the most basic version, the device ID is actually defined by the GenTL

2. The given ID can also be a user defined name. The user defined name is set

3. The serial number of the device can also be used as ID.

All three options can be seen in the output of gc_config -l.

Optional Interface ID prefix

If the given ID contains a colon (i.e. :), the part before the (first) colon is interpreted as interface ID and the part after the first colon is treated as device ID. This is the format that gc_config -l shows. A device with the given ID is only sought on the specified interface. This can be useful if there are several ways to reach a device from a host computer, e.g. via wireless and wired network connection, but a certain connection type (e.g. wired) is preferred due to higher bandwidth and lower latency.

Examples: eth0:00_14_2d_2c_6e_bb, eth1:02911931 or wlan0:rc_visard

A colon at the beginning of the ID effectively defines an empty interface ID which triggers looking on all interfaces.

If the given ID does not contain a colon, the ID is interpreted as the device ID itself and is sought throughout all interfaces as well.

Transport Layer

The communication to the device is done through a so called transport layer (i.e. GenTL producer version 1.5 or higher). This package provides and installs a default transport layer that implements the GigE Vision protocol for connecting to the Roboception rc_visard. According to the GenICam specification, the transport layer has the suffix '.cti'. The environment variable GENICAM_GENTL32_PATH (for 32 bit applications) or GENICAM_GENTL64_PATH (for 64 bit applications) must contain a list of paths that contain transport layers. All transport layers are provided as systems to the application.

For convenience, if the environment variable is not defined or empty, it is internally defined with the install path of the provided transport layer (as known at compile time!). If the package is not installed, the install path is changed after compilation or the package is moved to another location after installation, then the transport layer may not be found. In this case, the tools show an error that looks on 64 bit systems like:

In this case, the corresponding environment variable (see above) must be set to the directory in which the transport layer (i.e. file with suffix '.cti') resides.

Under Windows, as second fall back additionally to the install path, the directory of the executable is also added to the environment variable. Thus, the install directory can be moved, as long as the cti file stays in the same directory as the executable.

Network Optimization under Linux

When images are received at a lower rate than set/exepected the most likely problem is that this (user space) library cannot read the many UDP packets fast enough resulting in incomplete image buffers.

Test Script

The net_perf_check.sh script performs some simple checks and should be run while or after streaming images via GigE Vision.

./net_perf_check.sh --help

Jumbo Frames

First of all increasing the UDP packet size (using jubo frames) is strongly recommended! Increase the MTU of your network interface to 9000, e.g.

sudo ifconfig eth0 mtu 9000

Also make sure that all network devices/switches between your host and the sensor support this.

sysctl settings

There are several Linux sysctl options that can be modified to increase performance for the GigE Vision usecase.

These values can be changed during runtime with sysctl or written to /etc/sysctl.conf for persistence across reboots.


If the number of UDP RcvbufErrors increases while streaming, increasing the socket receive buffer size usually fixes the problem.

Check the RcvbufErrors with net_perf_check.sh or

netstat -us | grep RcvbufErrors

Increase max receive buffer size:

sudo sysctl -w net.core.rmem_max=33554432


Changing these values is usually not necessary, but can help if the kernel is already dropping packets.

Check with net_perf_check.sh and increase the values if needed:

sudo sysctl -w net.core.netdev_max_backlog=2000
sudo sysctl -w net.core.netdev_budget=600

2024-07-20 13:25