With 21st century technology, standard Ethernet can deliver on the highest performance demands of coordinated motion and integration of plant information
services. All on the same wire, at the same time.
Instead of managing all these networks, let's decrease that effort and focus more time on understanding and using the information that our network delivers.
In the industrial automation world, we have a network problem: our networks require too much time and money to design, install, and maintain. It's not that any
one network is unduly complex or time consuming. But the fact is; we have too many different networks. And that adds up.
Take any given industrial machine - there's a network to provide information to the user (normally Ethernet), a network to communicate with I/O and variable frequency
drives (usually Profibus or ControlNet), a network to communicate with simpler devices, such as sensors and control gear (usually ASi or DeviceNet), a network to
communicate with servos (usually SERCOS), a safety network (often SafetyBus P) and proprietary networks to synchronize controllers when motion must be distributed.
Each network requires different installers, training, software, tools, etc. The list of differences alone is huge.
So let's do something better with that effort. Instead of managing all these networks, let's decrease that effort and focus more time on understanding and using the
information that our network delivers.
Convergence is going to be the primary driver for cost savings in industrial automation over the next decade. The promise of industrial Ethernet is that multiple
networks will converge into one network, which can be installed once, requiring one set of training tools. But there are also requirements from the automation side:
put simply, if Ethernet does not deliver a significant step forward on today's motion solutions then something else will have to, even if it is based on Ethernet.
Convergence of DCS, PLC, drive controller and motion controller into the Programmable Automation Controller is one part of this. Convergence of I/O, sensor, actuator,
motion and safety networks into Ethernet will be another principal part. Convergence of information into a single database starting at the device and propagated through
the control system and into the supply chain is another. CIP Motion™ is one of a coordinated set of activities to help realize these benefits and to set them free to
build innovate solutions to problems, rather than worry about how the technology works.
Servo or induction motor?
In our current industrial environment, shifting effort to understanding and using the information that our network delivers can also be complex. Let's look at a simple
task: to move a part from point A to point B with a rotating shaft.
Should that shaft be connected to a servo or an induction motor? They are mechanically similar and a decision can be made on performance and fit. The same is true of
servo or variable frequency drives which are electrically similar. But then when we go back through the network to the controller we find that servo controls and frequency
controls use completely different languages and syntax; the profiles are different. The decision about which motor to connect to a shaft too often affects the skill-set
needed from the software application engineer!
But now we have ‘convergence'—the migration of all of these services onto a single network, Ethernet. This single network not only reduces installation and maintenance
cost but dramatically increases flexibility and adaptability of facilities. And these are all objectives that we strive for in the manufacturing space.
The determinism problem
Today's EtherNet/IP implementation of CIP Motion benchmarks 100 axes being coordinated with a 1 ms motion planner update to all axes. Further requirements include the ability
to coordinate an axis connected to its controller over SERCOS with an axis connected to its controller via Ethernet. Finally, the solution must allow for low performance
induction motors to be coordinated with high performance servo motors using the same profile. And all must be achieved without changing any of the four lower layers of Ethernet,
using standard, unmodified Ethernet and TCP/UDP/IP.
Two technologies have been developed in order to achieve this goal:
- CIP Sync™: IEEE1588 Precision Clock Synchronization mapped into the CIP object model; and
- CIP Motion: a set of application profiles designed to allow position, speed and torque loops to be closed within a drive (e.g. V/Hz, Vector and servo control type drives).
To understand how these two technologies work together we must first remember that there are two principle measurements in the calculation of a closed loop: the change (delta)
in position (or speed or torque) and the change (delta) in time. Inaccuracy in measurement of either results in an inaccurate calculation and unstable control.
Traditional networked motion control solutions use scheduled timeslots on the network to allow each device on the wire a "slot" - or schedule - in which to communicate with other
devices. As each drive or servo controller calculates its timing for regulatory control, the "delta-t" value used internally for this control function is determined from the
timing of delivery of data itself. Hence, unpredictable delivery of data results in an error in the "delta-t" term of the loop and poor motion performance. Opponents of the
use of Ethernet for motion control rightly state that unmodified Ethernet cannot allow sufficient predictability of delivery of the data in the system. Therefore, the delivery
of the data cannot be trusted as a foundational component for the regulation of precise motion or drive control.
To use unmodified Ethernet, this determinism problem must be taken out of the network infrastructure and placed into the end devices that are communicating. It is now easy to
see where clock synchronization of devices adds value. When a message is transmitted, the time of the transmission is encoded into the signal. Because the clocks in the end
devices are tightly synchronized, a small amount of jitter in receipt time of the message is unimportant, because the change in time can be accurately calculated.
This begs the question 'What level of time synchronization is needed?' And of course the answer is 'It depends!' For servo controlled systems, clock synchronization in the 100s
of nanoseconds is required. For lower performance variable frequency drives synchronization levels in the order of a few 10s of microseconds are acceptable.
On chip IEEE 1588
One of the attractions of IEEE1588 is that it has been designed as a scalable solution, deployable across many different network types. The algorithm can be either soft coded in
the lower levels of the Ethernet application level software, or it can be accelerated or hard coded to drive interrupt routines.
As CIP Sync maps IEEE1588 into industrial devices, this then sets the user free of adherence to a single time base. Different devices can use the time base that best suits their
application without being limited by either the fastest or slowest device on the subnet.
Experimentation has shown that the ‘soft' version of IEEE1588 can deliver approximately 5us synchronization, which is more than adequate for variable frequency drives. Of course
implementation of IEEE1588 in itself does not enable motion control. CIP Sync is needed to enable understanding of how time must be used within the device.
Simple calculations now show that the 100 axis/1 ms target for coordinated motion control across multiple controllers and multiple variable frequency and servo drives is achievable.
It's important to note that industrial automation and motion control are not the only applications crying out for precision clock synchronization. The fastest growing segment
of the global Ethernet market today is in voice services and voice switching. NTP (Network Time Protocol) will not provide the degree of accuracy necessary to reconstruct data
in highly distributed packet switch systems and GPS is often too costly to implement.
Similarly, as the test and measurement world transitions from IEEE-488 (or GPIB) to Ethernet, time synchronization and stamping of data over Ethernet is critical. Military
and power generation applications are also under way. Companies like Intel building IEEE1588 accelerator circuits into their communications chips to serve these markets.
An installed base of more than 1 billion Ethernet nodes ensures that Ethernet will continue to evolve. EtherNet/IP takes advantage of unmodified Ethernet and so is able to
take advantage of developments in standard unmodified Ethernet. The developments discussed above are very much 21st century developments. If EtherNet/IP had modified standard
Ethernet, then these developments would not have been available and the promise of motion control on unmodified Ethernet would not be deliverable.
This commitment to unmodified Ethernet also provides a clear migration path beyond 100MBps to 1GBps Ethernet and even 10GBps Ethernet, and so even higher axis count applications,
or more importantly, more valuable information on the network.
The cost savings of convergence in networks is now achievable and would in itself be a strong enough reason to select CIP Motion. However, we have also identified a need
to make software easier to use and to make servo and variable frequency drives truly interoperable.
The first key to this door is the scalable nature of IEEE1588 which allows lower performing devices to accept lower levels of synchronization accuracy and therefore greater
variability in communications and processor selection.
The second key is a drive profile (representation of a device that describes the device's data and behavior as viewed through a network) that is simultaneously complete
enough to allow for position speed and torque loops to be closed, while at the same time is flexible enough for configuration and parameterization of any type of drive.
The third and most critical key is a software tool where the motion path definition is completely abstracted from the drive definition of the axis. It is this abstraction
that allows the engineer to program and configure precise position and speed curves, neither knowing, nor caring, what type of drive and what type of motor is connected
to the rotating shaft.
This allows the motion control discipline and the variable frequency drive control discipline to converge. It could be argued that it even starts to allow overlap between
standard variable frequency control and servo control application. Overlap is good because without overlaps there are gaps, which result in inflexibility. Ultimately
the benefit of CIP Motion will be an end to applications where variable frequency drives or servo drives are misapplied simply because the software tools do not allow
them to work together.
CIP Motion has punched a hole in an urban myth of industrial automation. The myth: Standard Ethernet cannot be used for motion control and so something needs to be
developed that is based on Ethernet; adding time slice algorithms to the firmware or hardware of devices, or simply building a proprietary network with CAT5 cables
and a gateway to standard Ethernet.
This is important because even today users are starting to run up against limitations in their current implementations. They want coordinated motion control with
more than 32 axes on a single network. They want to synchronize axes connected to different SERCOS rings or connected to different controllers. They want a common
programming and configuration environment for variable frequency and servo drives. They want discrete inputs and outputs integrated with rotary machines. They
want video cameras mounted inside the machine. Many suppliers have developed proprietary technologies that address all of these issues but EtherNet/IP with CIP
Motion is the first to address them all using entirely standard Ethernet.
In order to deliver these user benefits, ODVA has enhanced and continues to enhance the CIP (Common Industrial Protocol) and EtherNet/IP specifications to
address four critical motion control requirements:
- Controller to drive interfaces;
- Controller to controller communications;
- Precision time synchronization between devices (or CIP Sync)
In order to deliver the benefits of Ethernet, there are also demands from outside of the ‘motion centric' world:
- Standard infrastructure components must be usable;
- Layers 1 and 2 of the Ethernet implementation of the ISO-OSI 7 layer model may not be changed;
- Existing standards should be used, not act as a basis for further development; and
- The solution must be applicable on standard off-the shelf chips.
There are also requirements from the application developer community:
- The network must be able to support 100 axes with a 1ms update rate;
- Position loops must be closable at a time base of less than 125 uS;
- Any axis can be either a variable frequency or servo drive with full networked configuration;
- It must be possible to synchronize discrete signals with a motion axis; and
- I/O, video and voice traffic must be carried on the same Ethernet as motion traffic.
These developments will not remove the status of motion control as a unique discipline in industrial automation. Instead, they will set free engineers to develop
more innovative and highly integrated solutions, increasing the level of user benefits delivered by control systems and machines. The network becomes simply a part
of the solution, not a part of the problem.
To learn more about how motion control is managed in the reference architecture to enable predictable system performance, sign up for the Industrial IP Advantage industrial network design training here.