Luminary Micro, in addition to providing some of the highest performance microcontrollers for motion control, such as the ARM Cortex M3, also supports IEEE 1588 and is one of the first motion control vendors to do so.
Real-time, digital control-system manufacturers generally improve response times with faster processors and internal bus structures to handle the integral data, calculations, and instructions. Although this approach has worked to improve numerous electronic intensive applications and broaden the fields in which these systems can be applied, it doesn’t address the motion control arena particularly well. In networked systems the fastest processors and backplanes are still challenged when it comes to making highly dynamic and precision motion-control systems.
In “hard, real-time” motion-control applications, the load becomes the primary consideration. It influences the basic system time constraints more than the control system’s ability to execute, according to the definition in “Measurement, Control, and Communication Using IEEE 1588,” by John C. Eidson. A control system’s ability to handle various dynamic loads creates some critical time-sensitive requirements for networks and controllers.
For example, controllers and networks often have to deal with just a few microseconds and at the same time insure that the results of certain critical commands reach their destination to perform the required function without a delay. In certain ac motor control algorithms, advancing or retarding the phase angle of the electrical current with respect to the motor’s rotor position has a dramatic impact on producing torque, and any network-inflicted delays could knock the system out of control. Consider a case where the control systems’ goal is to regulate current based on 1º angle resolution: A standard 1725 rpm ac motor control would have to be resolved into 10,350°/s. Based on the classical sampling theorem, the minimum sampling rate of two times the event rate is 21 kHz. This corresponds to approximately 50 µs just to define the rotor’s position with 1° accuracy. But when the motor is a small servo with an 8000-rpm top speed, then the time available to control the motor is divided by four, which decreases it to 12.5 microseconds.
The chart shows how processor utilization changes with PWM clock updates. Processor use is very efficient with respect to management of the motor.
Networks typically have trouble handling microsecond commands because they create significant latency due to the time needed to handle the information transferred between the main CPU and the axis controller. Latency comes from the time needed for two or more cycles to handle the information that must be sent and received plus the execution time required from both processors before position information can be returned. A fixed delay also appears between actual and commanded position regardless of the architecture implemented: It could be a backplane in a PC or PLC, an Ethernet interface, or some other high-speed connection. In low-speed or loosely coordinated motion systems this latency may not show up, but as the speed of the system increases, the latency cannot be diminished. It will begin to interfere, and it cannot be corrected.
Microprocessors and motion
Three basic building blocks are widely used for motion-control systems; microprocessors, DSPs, and microcontrollers. Each has its own unique advantages and disadvantages.
The Luminary Micro Reference Design Kits include free source code, schematics and a Graphical User Interface to facilitate monitoring the application and making code changes more visible.
Microprocessors: Microprocessor chips are relatively inexpensive and are widely used for numerous types of control systems. But with a 50 µs or smaller window commonly encountered for motion-control systems in the real world, the operating demands placed on microprocessors are often more than they can handle. Higher clock speeds and more efficient execution methods are increasingly forcing designers to find control chip approaches other than microprocessors for certain applications.
DSPs: One alternative often used is digital signal processors (DSP’s). They have excellent math processing capability, but don’t function well in multi-tasking or context-switching applications. The processor must halt execution of the main task in order to service the communications request, which forces programmers to work around the limitation. Interrupt Service Routines (ISR’s), which are critical path instructions may take between 20 and 60 clock cycles to execute, often too long for highly dynamic systems.
Microcontrollers: Another approach is a microcontroller. One primary distinction between microcontrollers and other processors is that its memory is part of the processor, not external to it. The same architectural problem exists in PC’s, and stacking the memory outside of the main processor creates significant delays to the system’s throughput. To circumvent this problem, an MCU (microcontroller unit) designed for motion-control systems by Luminary Micro, Austin, Texas, integrates memory, processing, and network support all in the same chip. This approach eliminates extra parts and contributes to a higher throughput than competing approaches. For
instance, ISR instructions that take 20 to 60 cycles to execute on DSPs require only 12 clock cycles or less on the Luminary Micro parts.
The main application runs in the foreground or base level while maintaining the motor control and communications tasking as needed.
Overall performance comparisons aren’t easy to make with this special MCU because the standard unit of measurement, MIPS (millions of instructions per second), does not take into account the relative efficiency of its instruction set. Moreover, this microcontroller’s higher efficiency lets it process the main application, interrupts, and communications tasks without losing motor synchronism. Somewhat like “multi-threading” software, the microcontroller is able to keep track of the different processes simultaneously.
Many motion control systems require two processors and complex programming techniques to factor the application. This often leads to inefficient use of processor resources.
Networks and motion
The second major building block of a modern motion-control system is its network, and perhaps not surprisingly, Ethernet is often the first type to be considered. It is widely used in Local Area Networks and has become the standard for office and consumer applications. Continual increases in speed from 10M, 100M, and now 1Gigabit has led to Ethernet’s overwhelming acceptance, which has helped reduce costs dramatically. As a measure of how widespread Ethernet has become, Ethernet connectors, cable and connection tools are available at most computer, electronics, hardware, and consumer appliance stores.
Consumer-type Ethernet may be great for office and home use, but it has problems for industrial control. For example,determinism, the predictability that a message arrives, is not a feature of Ethernet in office and consumer applications. A failed message can be rebroadcast until it finds its destination, but the system has no provision for determining the precise time it arrives, the most critical factor for motion control systems that depend on microsecond timing.
Fortunately, however, Ethernet now contains several improvements and a few added features that insure network determinism. Vendors and standards groups made amendments to take advantage of the low cost of Ethernet components. For example, EtherCAT and Ethernet POWERLINK are two more recent permutations of the basic Ethernet platform that add determinism.
Another method of improving Ethernet is making time slots to carry messages. In fact, this is the basis of the Sercos network for motion control. One of its unique features is its fiber optic physical layer, which inherently eliminates electrical noise.
In November of 2002, after several years of development and refinement, the IEEE 1588 protocol was published, which adds features to Ethernet that guarantee time synchronization across the local area network. The two major features are a “grandmaster clock” that synchronizes all clocks in all the devices on the local area network, and a device ID for each device in the network. Non-synchronized devices can co-exist, but synchronized devices communicate with absolute determinism and granularity down to a single cycle of the grandmaster clock.
Finally, in an effort to move Ethernet into the industrial mainstream for control, the Open DeviceNet Vendors Association (ODVA) supports EtherNet/IP as an extension to consumer-grade Ethernet. The improved Common Industrial Protocol (CIP) is intended to address determinism and time synchronous behavior in order to solve an overall network problem and bring industry’s legacy networks into one solution. The technical elements of CIP are based on IEEE 1588 protocol.
This combination of an improved MCU controller targeted for motion-control systems and the work undertaken by standards groups to improve Ethernet have proved that the combination is more than ready – and indeed is available for motion control systems.
Filed Under: Electronics • electrical, Ethernet — cables • hubs • switches, Motion control • motor controls