Systems
engineers are frequently tasked with determining the appropriate
network for optimizing the performance of their motion control
system. Usually, the two competing protocols are TCP/IP over
Ethernet or a Fieldbus architecture such as CanOPEN, SERCOS,
or DeviceNet. Such Fieldbus networks rely on small, deterministic
packets of data that are transferred between a single master
and multiple slaves. This method works very well when the
information is repetitive, short, and required at regular
intervals. The
data could include streaming position commands for each axis
of motion.
Ethernet
protocol provides the engineer with a flexible, scalable network,
where large packet size and collision recovery allow intelligent
devices to seamlessly communicate with one another. Since
any device with an Ethernet network can generate messages,
the potential for a collision exists.TCP/IP
recovers from a collision by re-transmitting the packet after
a small, randomly generated delay.
This
effect leads to non-determinism of less than 1 millisecond.
However, because all the servo loop and coordinated motion
profiles occur off-network, multi-axis controllers are insensitive
to non-determinism. In addition, with the lighter network
traffic, the chance of a collision occurring at all is far
less likely than with a fully loaded network.
Galil
Motion Control has developed the flexible-distributed network,
(see scenario 3), which lets the user configure the system
so that axes that are in close physical proximity to each
other are controlled by a multi-axis motion controller. Also,
each of these multi-axis nodes can, in turn, communicate with
a master controller or the host PC. This keeps all time-critical
operations such as servo loop closure, coordinated motion,
and I/O handling local to the node. Plus, the network is used
to merely transmit high-level supervisory commands, such as
“Execute a process” or “Tell the master the
current I/O status”.
Because
of the flexibility of TCP/IP packets, Ethernet is the network
protocol of choice. Because of the vast popularity of the
Internet, almost all PC users are familiar with some form
of TCP/IP. By relying on the local loop closure and motion
profiling inherent to all Galil motion controllers, a systems
engineer can create a robust control network with off-the-shelf
commercial products and a user-friendly, popular, open-source
serial data protocol.
A thorough
examination of the various protocols is given in Application
Bulletin #5449 (pdf format, 55k).
|
The
following diagrams detail three control schemes that
the designer may choose – each with benefits and
drawbacks. Here, the network depends on the devices
in the system.
 Scenario
1
In
this scheme, a host PC sends high-level commands to
a centralized controller, where all loop closure, coordinated
motion, and I/O handling are performed on the controller.
This architecture is best suited for either Ethernet
or RS-232.
Scenario
2 This scheme shows a host PC sending real-time position
trajectory commands to a network consisting of several
single-axis drives. This data is regular and repetitive.
If the system requires loop closure such as SERCOS to
occur over the network, then a deterministic network
is necessary. If the system contains intelligent drives
capable of closing the servo loop internally, then either
a Fieldbus or Ethernet network is acceptable.
 Scenario
3
Here
a host PC sends high-level motion commands to a network
consisting of
multi-axis motion controllers and I/O. In addition,
text messages and irregular (unsolicited) data will
be transmitted from one device to another. All servo
loop closure, coordinated motion, and I/O handling is
performed inside the controller. For this architecture,
Ethernet is the best option. |
|
