Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Embedded

Improve systems' real-time performance

Posted: 16 Nov 2007 ?? ?Print Version ?Bookmark and Share

Keywords:multicontext architecture for industrial system? real-time industrial system design? flexible embedded processor?

Industrial applications have more inherent challenges than consumer applications. They must operate within very tight time constraints and require a high degree of deterministic predictability. Typically they are deeply embedded within production-floor systems, which are often in harsh factory environments. Manufacturing lines require precise sequencing and synchronization of operations that, in turn, demand consistent real-time performance from embedded controllers.

Industrial automation controllers need to deliver deterministic real-time performance. They must also be compatible with the surrounding production environment and various communications networks. Many embedded industrial applications are built around field buses, which are optimized to provide deterministic performance for controlling runtime machine operations. However, there is a growing movement toward the use of communications standards such as Ethernet for intermachine and production-wide networking.

The inherent differences between deterministic field buses and standards-based networks can make it particularly difficult to integrate these disparate communications protocols using conventional microprocessors. In addition, the need to support a range of communications interfaces can lead to increased inventory and logistics costs if different components are needed for each interface.

Most microprocessors spend much time handling switching and managing tasks. And commercial microprocessors don't distinguish between real-time and safety-critical tasks vs. user application code and user-interface tasks. So it becomes almost impossible to ensure deterministic low-latency processing of real-time and safety-critical tasks without additional software (i.e. RTOS). Even then, the designer has relatively little control over how the RTOS manages task switching and prioritizing of code execution.

Multicontext architecture
A new chip-level architectural approach addresses these shortcomings by using multiple independent hardware contexts and by handling many of the real-time tasks within the chip itself. This architecture allows for hardware-level partitioning of management functions and real-time task execution. It allocates processing priorities across as many as five separate hardware contexts, thus enabling the highest-priority tasks to always have immediate and unblocked access to processing resources.

For industrial applications, the ability to control latency is critical. It enables tasks such as sensor reading, control metrics, motion control and PLC outputs to be executed immediately and predictably. By dedicating code such as real-time loops to a specific hardware context within the chip, the multicontext architecture can minimize latency.

Similarly, for situations that require interfacing with multiple field buses or standards-based networking, multiple hardware contexts provide an excellent way of isolating the differences between the interfaces. A deterministic bus such as CAN, Modbus or Real-Time Ethernet can run at full speed on a dedicated hardware context without having to wait for processing resources the way it would in a single-MCU implementation.

Because real-time tasks are handled by independent hardware contexts, the multicontext architectural approach also means that much of the software overhead associated with real-time operation can be eliminated. The simplification of software development represents a major cost and time-to-market savings, and consequently, more robust systems.

Another benefit of moving major processing activities to separate hardware contexts is the reduction or elimination of saving and restoring states when switching between tasks. This avoids the requirement to push and pop registers onto a stack.

Another major advantage comes in the memory management arena. The context-aware memory protection unit manages memory partitioning for the different contexts without additional software overhead. It automatically associates specific blocks of memory with specific hardware contexts, assigning appropriate permissions at boot time.

Debug is also simplified by the new processor architecture. With industrial systems often embedded deeply within customers' production-floor applications, there is limited room for engineers to set up monitoring equipment for in-situ performance measurement and debug. With physical access constrained, the ability to remotely control and analyze multiple facets of the machine can be an important advantage. Problems can then be quickly solved without disrupting the production environment.

The multicontext architecture allows for hardware-level partitioning of management functions and real-time task execution.

The multicontext architecture provides a high level of remote control and observability through the JTAG port. Because the emulation is context-aware, register sets can be independently switched in and out depending on the context. Simple debug tools allow for changing the register state or peripherals one bit at a time while clocking through the process.

Trace buffers for event logging can be allocated anywhere within the address space or even sent out over the Ethernet port. For a properly designed system, this means an engineer can easily define watch points and breakpoints, set the system to run and then view results remotely, without having to travel to or interfere with the customer's field operation.

A major challenge for designers is dealing with the relatively long life cycles of industrial equipment, especially against the backdrop of shorter life cycles for many of the processors used for embedded-control functions. Mainstream microprocessor providers must consider the volume of commercial mass-market opportunities as their key business drivers when making decisions about engineering changes, production quantities and end-of-life phaseouts.

Another factor accelerating the phaseout of many devices is the need to comply with environmental regulations such as the EU's RoHS. In many cases, suppliers are finding that the simplest alternative for older, non-compliant components is to discontinue them.

Implementing flexibility
To drive a new era of system performance that bridges both legacy requirements and emerging technologies, designers of industrial applications need a high degree of real-time performance and implementation flexibility from their embedded-processor choices. It is not enough to drive higher clock rates and more raw performance from traditional single-threaded MCU architectures. Current industrial apps require much greater flexibility at the chip level for partitioning and optimizing real-time functions. Designers also need relief from the burden of extensive software development efforts that have been required to overcome the limitations of conventional mainstream embedded processors.

By looking to new embedded-processor designs that provide simultaneous multiple hardware contexts along with integrated software library elements and debug capabilities, industrial-system designers are able to separate key real-time processes from less time-critical code--at the lowest level of implementation in the hardware processor itself.

The bottom line for both industrial-system developers and end users will be new levels of performance, modularity, interoperability and extensibility for the next-generation industrial applications.

- Jordon Woods
Co-founder and Chief Technical Officer
Innovasic Semiconductor

Article Comments - Improve systems' real-time performa...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top