Heart of the Network

Microprocessors and Software Are the Underlying Technologies Driving Advances in Industrial Networking. Here's How You Benefit From the Evolution

What gets your blood flowing? If you're an industrial network, then cables and wires are the arteries, electricity is the fuel, and digital data are the nerve impulses from the enterprising brain.

So, what centrally located device pushes and coordinates the rhythms of this circulatory network and the nervous system it supports? Where's the heart?

Similar to so many critical, taken-for-granted workhorses, it's hiding in plain sight. The heart of any industrial network is, of course, the central processing units (CPUs) in its equipment and subsystems. These ever-faster, more powerful, less costly and more varied microprocessors perform almost infinitely more calculations, run more sophisticated software, and exchange information over networks managed by Ethernet switches and routers, which have their own onboard chips.

It's no stretch to say that microprocessors have helped industrial connectivity to evolve over the past 25 years from hardwiring to fieldbuses to Ethernet to wireless, and onward to Internet-enabled monitoring and the recent emergence of cloud-based services and virtualized computing on fewer servers.

"When you think about it, microprocessors and software allow all parts of our networks to function," says Corey Heckman, automation manager at Control Systems 21, a system integrator in Dillsburg, Pa. "Without them, we'd have to try to use relays, big logic panels and hardwiring. It would be a nightmare because all of our projects would be unworkable."

Control Systems 21 implements SCADA and PC-based devices for process control systems, mostly in industrial and municipal water/wastewater systems. To bring in required signals and data, its networks transition rapidly from wired to wireless hardware and back again, and their communications methods also move from serial to wireless protocols to virtual private networks (VPNs). As a result, Control Systems 21's often far-flung networks use serial and Ethernet radios, cellular communications, and other wireless devices.

For example, for a mid-sized city to add wastewater metering to its municipal water system, Heckman's team is networking all of the system's pumps, PLCs, digital video recorders (DVRs) and other components at about 25 different sites. These locations include wells with high-level monitors, inputs for water from other systems, a reservoir with booster pumps, and data-collection metering for the wastewater system. The main network is using Modbus TCP/IP with Ethernet switches, but its communications also use the Internet for remote video surveillance of stations, VPN tunneling, Phoenix Contact's Trusted Wireless Ethernet (TWE) radios and other components. All of these devices are polled by PCs running Trihedral's VTS SCADA/HMI software. Data from the remote stations comes in via serial UHF licensed or 900 MHz unlicensed radios or VPN tunnels, and an OPC server translates and coordinates this input, so the SCADA/HMI software can analyze and display it.

"I was even able to set up an app, so the city's engineers can monitor the water/wastewater system with Android smartphones," Heckman explains. "Of course, this web-based application also has verification steps and security certificates to make sure that access is secure."

Cellular communications weren't used much on the plant floor in the past, but that's changing as bandwidth increases through the growth of 3G and 4G technologies, reports Rich Harwell, connectivity manager for Eaton's controls and automation group. "Many web-based applications and access to local area networks (LANs) are being enabled by cellular, and so it's going to show up in more machines and industrial settings, too," he says.

How Chips Evolve and Help

Although microprocessors have multiplied their capacities and speed many times over, while shrinking in size and cost, it can be harder to understand exactly what this means for industrial networks and their users and applications. Who can really comprehend billions and trillions of calculations flashing along in a microscopic space?

Certainly, the end result and evidence is that useful tools such as Ethernet switches have been popping up, but learning more about microprocessors' internal workings could help convince more potential users of their capabilities and reliability. One often-repeated rule of thumb is that anyone planning a typically longer-lifecycle industrial network should make sure to install enough bandwidth, maybe even Gigabit Ethernet, to handle all the traffic from IT components, mainstream electronics and other devices with shorter lifecycles.

"Microprocessors might be on the lowest layer of the network, but they have a huge influence," says Mike Justice, president of Grid Connect, which manufactures networking products. "For example, when I was at Lantronix about 10 years ago, we introduced X-Port, which was just a simple RJ-45 connector with an Ethernet chip inside. However, it let any serial device communicate via Ethernet, and so GE used it to add multiple networks to control devices for elevators, pneumatic tubes and other applications."

Last month, Grid Connect introduced its gridARM microprocessor architecture to enable 1 Gbps Ethernet at low cost in scales, barcode readers, I/O points, and web-based controllers. "Lots of office devices use Gigabit Ethernet, and so the factory will need it soon to be scalable for the future," Justice adds.

While microprocessors have shrunk and speeded up and done the same for all the devices in which they're embedded, the persistent needs of many applications and users have also caused those chips to adapt and adopt more varied functions. For example, older-generation semiconductors and standard microprocessors were cheap and plentiful, but they didn't have the design flexibility and functions that many device builders needed. So they gave way in many cases to application-specific integrated circuits (ASICs) that were produced in lower volumes by third-party fabricators. That gave users the design-stage flexibility they needed, even though they were typically built to do just one job. More recently developed were field-programmable gate arrays (FPGAs), which are similar to ASICs, but can be programmed by the user for specific tasks after they're manufactured. And, of course, even more flexibility and capacity has been added since then to FPGAs and other systems on a chip (SoCs).

"Because so many control and automation applications need determinism, the Ethernet-based media access controller (MAC) layer in FPGAs and other processors can be modified to add more timing blocks to achieve the determinism required," says Jason Chiang, senior strategic and technical marketing manager at Altera. "This is the beauty of FPGAs. At the physical, printed-circuit-board (PCB) level, you can use one card to run multiple fieldbus protocols, such as Profinet, EtherNet/IP or SERCOS III. So we've been seeing a shift to more programmable logic and industrial networking on FPGAs. We make these for customers like Rockwell Automation and Siemens, and then they add their secret sauce to the Ethernet MAC hardware and run their software stack on it."

Altera makes microprocessors, such as its Cyclone IV SoC FPGA, and industrial networking kit (INK) for developers (Figure 1). In mid-October, it also announced a new class of SoC FPGA with a full ARM core processor on it.

We're likely going to see three or four times the amount of Ethernet we have now because the chips and software are fast and cheap enough to handle it, and because they can run over copper, fiber or wireless, depending on the application's environmental needs," says Mike Miclot, marketing vice president at Belden. "This will mean more real-time control and safety integrity level (SIL) functions at the same time, SCADA combined with security and surveillance, and video streaming in a lot more applications."

This past January, Advanced Micro Devices (AMD) introduced an embedded G-Series Accelerated Processing Unit (APU), which is small and fast enough to operate at 5 W in a headless, embedded PC form factor that runs Linux and can accept high-definition video (Figure 2). It can network everything from sensors to robots via Ethernet, monitor and network plant-floor operations, effectively replace many PLCs, and even has virtualization hooks to optimize the performance of a virtual operating system or "hypervisor," according to Dave Jessell, AMD's embedded enterprise development manager. "Software usually gets most of the limelight, but it's the hardware standards these microprocessors are based on that provide the foundation for software developers to do industrial applications," Jessell explains. "G-Series APU is able to calculate algorithms closer to real time, which allows it to achieve better control and efficiency, and connect many industrial devices as thin clients to a virtual server."

Many machine builders demand that their control and automation suppliers provide components with more-open networking, so the end users of their machines won't be constrained by proprietary hurdles, adds Bob Ferrar, director of the intelligent systems group at Intel. "As a result, some suppliers are moving from proprietary and non-standard ASICs to more open ones based on standards like IEEE 1588 and others," he says. "Sometimes, PLCs also can't talk to each other very well, and so they might need an FPGA with protocols loaded on it to translate between them. However, this can cause speed to become an issue, which might be acceptable for an application cycling at 500 ms, but not for a robot arm running at 50 ms."

Simplify and Combine Networks

Just as more sophisticated software in the background allows users to type in less code and do more point-and-click programming, more capable and widely distributed CPUs are simplifying plant-floor systems and networks, and making it easier to use and service them.

For instance, to coordinate the data and control needs of an expanding woodyard operation at Packaging Corp. of America (PCA) in Valdosta, Ga., its two system integrators recently decided to combine two separate control systems (Figure 3). This operation includes a woodchip stacker-reclaimer from Bruks Rockwood, which uses Allen-Bradley controls and communicates via EtherNet/IP. PCA works with system integrator Electric Machine Control (EMC) in Birmingham, Ala. However, the yard's log-handling cranes, debarkers and conveying systems are supplied by Fulghum Industries in Wadley, Ga., which employs other controls, including PACs from Opto 22. This equipment is integrated by Advanced Control Solutions (ACS) in Marietta, Ga.

To interconnect the two control systems for better data throughput and coordination of operations, the two integrators decided to maintain Opto 22's local, distributed control system, and use the Valdosta plant's Ethernet network. The Opto PAC could talk directly to the A-B PLCs because Opto 22 recently added support for EtherNet/IP.

"Once enabled, our I/O can be added to Logix platforms, and communicate with PLCs with no programming required," says James Davis, Opto 22's senior application engineer. "Also, Opto 22 controllers can serve as slave devices or adapters in the Logix architecture." Likewise, ACS's systems engineer, Sean O'Rourke, worked with EMC to interface to the A-B PLCs via fiberoptic connections, which provided a reliable, high-bandwidth, high-noise-immunity, long-distance physical network.

"Configuration and setup was simple," O'Rourke says. "We only needed to define the assembly instances, assign inputs or outputs, and specify the number of bits for how long each instance was going to be."

This configuration was then downloaded to the Opto PAC, and all that remained was to configure the A-B RSLogix software, and define communication to the PAC as a "generic Ethernet module." The two integrators add that reconciling two disparate control systems is becoming common because end users want more networking options as they build or modify their controls.
Similarly, one of the main benefits of smaller and cheaper microprocessors is that they allow more sensing and data collection points and increasingly wireless networking and control components, which means more can be installed in more places, perform more monitoring and data acquisition, enable better decisions, and do more track-and-trace documentation, especially in pharmaceutical and food and beverage applications.

"There are far more levels of visibility possible now by the plant or the administration," says Jeff Smith, senior engineer for global controls architecture and manufacturing networks at American Axle & Manufacturing (AAM) in Detroit. "As a result, we can zoom into the equipment in a plant in Brazil or anywhere worldwide, and see how well they're following established metrics, or we can drill into PLCs on a machine anywhere and troubleshoot it."

Brian Oulton, Rockwell Automation's networks marketing director, adds, "People are learning that they don't always need standard PCs, but can use super-lightweight client devices instead. This is one of the biggest results of everyone adopting standard unmodified Ethernet and IEEE 802.11n wireless, so they can use more commercial technologies in industrial settings. Likewise, I believe we're also sitting on the edge of real-time data playing a much bigger role in manufacturing, and then being able to use and respond to it much closer to real time."

On the Internet, In the Cloud

While some credit CPUs and software with driving industrial network performance, others believe the pull of Internet Protocol (IP) is a primary enabling force. And, of course, it's even more attractive now that the cloud and other Internet-based services have arrived.

"It's true that IP needs processing hardware, but it provides ubiquitous communications that are well-established worldwide and have many options," says Benson Hougland, marketing vice president at Opto 22. "However, industrial networks and IP have to interact with the real world at some point, and so the PHY form interface is what works with the microprocessors at low power and in combination with firmware or software to make those communications possible. This is why we evaluate new technology, provide ASICs at the I/O level to localize processes, and distribute intelligence and computing wherever it's most effective. In general, the combination of data processing and communications is making more devices autonomous, and so we have more mobile applications and the cloud."

Cloud-based services could further reduce data processing costs for users, including those in industrial control and automation, Hougland adds. "Using a virtualized server somewhere and file synching software means you don't have to build and house your own servers."

Eaton's Harwell adds, "We're seeing more fully web-enabled devices at the machine level, which reflects the increased connections between machines and the continuing growth of industrial Ethernet, which is itself a result of less costly, faster and more powerful data processing. Many machine builders are wrapping their arms around this because, instead of having to add a traditional SCADA package, they simply can build a regular operator interface at the machine level, and send data via a simple Ethernet port. They even can include real-time data views, data storage, archiving and analysis function, so they don't need a data acquisition (DAQ) unit either. "

Virtualization Is for Real

Probably the spookiest change caused by faster and stronger microprocessors is their ability to use leftover computing capacity to collapse and combine several former operating systems or PCs, and run them as virtual functions on one or two physical processors in servers that can be located pretty much anywhere.

"In PLCs across the board, we see previously separate functions and components collapsing down into individual devices that can run them all," says Intel's Ferrar. "For instance, a user interface can run on one virtual operating system, while some control functions can run another virtual real-time operating system (RTOS), but they're both on the same physical core."

To make sure virtual computing is carried out securely, Trusted Execution Technology (TET) modules are being developed and added to Intel's own CPU architecture to help virtual computing applications check and make sure they're operating securely, and that no other communications are being allowed, Ferrar says.

Likewise, running virtualized computing on just one or a few servers instead of many hardware PCs can make it easier to recover from accidents and possible security breaches. "If you don't depend as much on physical hardware, then you can recover in a few seconds or minutes simply by reverting to a safe version of your virtual computer, rather than waiting days or weeks for hardware computers to be checked and fixed," says Kevin Staggs, engineering fellow for cybersecurity research at Honeywell Process Solutions.

Besides advising process engineers to adopt IT-based security practices, some microprocessors tack on security functions, such as setting stricter requirements for executing tasks, building protection layers into their silicon, and establishing chip-level encryption and firewalls, Staggs adds. Trusted Computing Group's Trusted Platform Module (TPM) is a microcontroller that can securely store artifacts used to authenticate its PC's platforms, such as passwords, certificates or encryption keys.

Surprisingly, just as hardware and their applications are condensing due to faster, higher-capacity data processing, many control and IT engineers not only cooperate more, but are beginning to see their job descriptions merge and their organizational supervisors become the same person. "We now have a number of folks who are bilingual in controls and IT," says AAM's Smith. "It helps to switch controls people on the plant floor with IT people and with administrators for a while, so they can learn about each other's positions."