While still very much a mechanical environment, the modern factory with its increasing automation increasingly relies on advanced microelectronics. The Industrial Internet of Things (IIoT), which is fundamental to what most call the fourth industrial revolution, could not exist without chips.

Consider large, powerful machines forging tons of molten metal controlled by transistors 100,000 times smaller than the width of a human hair. Although the contrast is stark, the transistors are in every sense machines, too. In their case, however, their raw material is data.

IIoT Brings Data Integration

Industrial operations are invariably high-speed ones, which puts pressure on the underlying control technologies. Data acquisition and analysis are now running in parallel with real-time control, prompting OEMs to increasingly invest in new technologies to bring IoT to an industrial environment.

Machine builders are integrating industrial PCs into their products to deliver new features and value. The industrial PC combines industrial control with data acquisition and wide-area connectivity, enabling machines to be guided by a broad range of information and managers to better assess performance. The industrial PC often displaces simpler legacy automation solutions like programmable logic controllers (PLCs).

Data gathered in the industrial environment and shared with cloud-based services enables large-scale analysis of process quality, productivity, maintenance requirements, material consumption, waste, workplace safety, and other conditions. Distributed sensors generate the raw data, which is then analyzed to create actionable insights. These insights drive higher productivity through gains in efficiency.

While the industrial PC is compatible with cloud services, edge computing is a compelling alternative, bringing high-performance computing directly to the factory floor. Data is analyzed as it’s generated at the locus of the activity. Why take time to send it through a network when you don’t need to and can’t afford to wait? Edge computing is supported by high-speed connections based on time-sensitive networking (TSN), a new technology that uses Ethernet to reduce latency and enable deterministic delivery.

Solutions for All OEMs

Introducing new technologies like edge computing always carries some risk. The advantage goes to larger OEMs who can better afford to make a risky investment and are therefore first to realize new productivity benefits. Smaller manufacturers, after seeing concepts proved, are forced to play catchup.

So not surprisingly, machine builders with large manufacturing customers introduce new technologies as soon as they can. Smaller OEMs may still implement their automation in house on a smaller, simpler scale, often replacing their PLCs at a slower rate, if at all.

Fortunately, PLCs can be made smarter. Today, more PLCs have network connections and embedded internet technologies, turning them into tiny web servers (though not really parts of the IIoT). The micro-PLC, rather, is a step in the IIoT direction for smaller manufacturers, serving as a gateway to bring facilities online and offload some control tasks. Many micro-PLCs are based on low-cost single-board computers, making them accessible to most.

There are a variety of viable approaches to smarter industrial automation. Manufacturers may choose a network of industrial PCs to control an entire facility. These OEMs may prefer a distributed system using cell-based edge computing while others may opt for micro-PLCs to offload local tasks while adding remote access. Each option has merit and relies on semiconductor technology.

Another area of industrial automation where semiconductor technologies are essential is machine vision. Augmented with artificial intelligence (AI) and machine learning, machine vision plays a significant role in accelerating automation.

Semiconductor manufacturers are developing and integrating support for AI at the chip level. This includes larger, more powerful cores with dedicated hardware accelerators. High-performance field programmable gate arrays (FPGAs) also support AI integration into industrial controls.

Inside an Industrial PC

As with any personal computer, the industrial PC is based on a high-performance microprocessor. Historically, the architecture of these processors has been based on the x86 instruction set, dominated by the two semiconductor giants, AMD and Intel.

Most industrial PCs are still based on an x86 processor running a Windows operating system. The PC may take various shapes, often housed in a rugged enclosure to protect it in a harsh operating environment. The motherboard will normally conform to industry standard formats, such as ATX, Micro-ATX, Mini-ITX, Pico-ITX, or Thin-ITX.

With microprocessor performance increasing, the physical size of the motherboard format is trending smaller. As this happens, we move from motherboard to embedded compute boards, to system-on-module (SoM) and as far as system-on-chip (SoC). Shifting from industrial motherboard to embedded compute board introduces a new portfolio of standards.

These standards include COM-HPC and COM Express from PCI Compute Manufacturers Group (PICMG), and from the Standardization Group for Embedded Technologies (SGeT) we get SMARC, Qseven and Open Standard Module (OSM).

Cooling is one limiting factor in the trend toward smaller form factors. The physical size of the processor relates to its compute capabilities. The preference in industrial applications is for PCs and boards that do not require forced-air cooling. Conduction cooling via a heat-sink through to the enclosure reduces size, power, and noise.

High-end microprocessors, such as those based on Intel’s 12th generation core, may feature four processing cores with additional smaller cores, and dissipate over 100 W of heat. Heatsinks in contact with a metal enclosure dissipate the heat generated by the processor, but this will impact the overall size and design of the enclosure.

Windows on Arm

The industrial PC combines industrial control with data acquisition and wide-area connectivity, enabling machines to be guided by a broad range of information and managers to better assess performance. (Image: Avnet)

Fortunately, low-power processors can often deliver the performance industrial applications need without generating as much heat. This approach works in deeply embedded industrial applications, particularly when using real-time operating systems. The Arm architecture is a popular choice for embedded compute boards, though a conflict in requirements arises in IIoT applications that need to run Windows.

To address this conflict, one can use an x86 processor in a fanless configuration and monitor the performance to remain within safe operating temperatures, or to emulate Windows on another processing architecture, such as Arm.

Emulation, however, often comes with limitations in compatibility and performance. Not all software written for Windows will run in emulation, and the extra resources needed to run emulation can limit its applicability.

Fortunately, Microsoft recently opened the core technology behind the Windows 10 IoT Enterprise operating system. Approved semiconductor partners are porting Windows directly to the Arm instruction set architecture. NXP Semiconductors is one of those suppliers approved to port Windows.

Augmented with AI and machine learning, machine vision plays a significant role in accelerating automation. (Image: Avnet)

Windows 10 IoT Enterprise now runs natively, without emulation, on the i.MX 8M Plus application processor from NXP. Avnet Embedded has developed a platform compliant with a SMARC embedded compute module based on the i.MX 8M Plus. Thus, the platform runs Windows 10 IoT Enterprise natively. Any application written for Windows 10 IoT Enterprise will execute on the Avnet Embedded platform. Avnet Silica and Witekio worked with Avnet Embedded to develop the platform and the board support package to port the operating system.

The platform delivers high performance with low power for industrial applications running Windows on a compact and energy-optimized hardware solution.

High-Performance Silicon for Industrial Applications

Further down the integration line we come to the system-on-module, or SoM. Beyond that we have the system-on-chip, or SoC. These systems invariably feature a microprocessor, but they may also include one or more microcontrollers. The microcontroller, or MCU, is a flexible device, designed for interfacing to other hardware devices such as power transistors that energize a motor’s coils. MCUs designed for digital motor control incorporate a processing core, executing algorithms tuned to combine performance and efficiency.

The MSC OSM-MF-IMX8NANO is a SoM designed for soldering onto a carrier board, as any other component would be. This means there is no connector, which reduces size, lowers cost, and improves heat conduction. The module is based on the i.MX 8M Nano application processor from NXP. Configuration options include a single, dual, or quad-core Arm Cortex-A53, an Arm Cortex-M7 real-time processor, and a multimedia graphics processor. The OSM size M module measures 45 mm by 30 mm, but the OSM standard includes size S, measuring just 30 mm by 30 mm.

An SoC is a single device with a very high level of integration. In theory, an SoC could replace an industrial PC in some applications. Design flexibility comes from using an SoC that includes an FPGA fabric. FPGAs have long been used in high-performance applications thanks to their high speed. The configurable logic implements functions at a transistor level in hardware rather than in software.

The trend in FPGA-based SoCs is to combine hardwired functions for maximum performance, alongside configurable gate array logic for custom hardware, and complemented by processor cores to run standard software.

The PolarFire SoC from Microchip is an example. The single-chip device integrates five RISC-V processing cores, an FPGA fabric, high-speed memory, high-performance communication transceivers, and general-purpose I/O.

Future Developments in Industrial Automation

Whatever approach you select, microelectronics play a critical role in industry, and one that’s not always visible at first glance. Consider: The industrial sector measures efficiency in terms of unit cost against total volume. Marginal gains at the unit level translate to big savings at the right volume. High-volume production delivers a lower unit cost, but it can have a correspondingly high environmental impact.

There is a trend toward lower volume and on-demand manufacturing to address overproduction and waste. This trend is enabled by new technologies that support the smart factory. A smart factory can adapt faster without sacrificing the efficiencies that normally only come with high-volume manufacturing.

A smart factory can adapt faster without sacrificing the efficiencies that normally only come with high-volume manufacturing. (Image: Avnet)

Moving to a smarter manufacturing industry requires new technologies, including initiatives such as TSN for using Ethernet in real-time control applications.

New developments at the silicon level include more powerful SoCs designed for harsh industrial environments. Modules and compute boards based on SoCs and SoMs provide a fast and standard way to integrate advanced processing.

Edge computing will continue to develop by adding AI and machine learning capabilities at the manufacturing cell. This will promote a smarter factory, one able to react quickly and efficiently to change.

Industrial PCs are spearheading the adoption of IIoT and continue to evolve. New form factors, higher performance, and lower power processors will support these developments.

Avnet is actively working across all these technologies to help customers adopt the solutions that are right for them.

This article was written by Philip Ling, Technologist, Avnet (Phoenix, AZ). For more information, visit here  .