The Internet of Things has moved from being a vague concept five years ago to something far more concrete and many of the articles in this year’s edition of Outlook underline this. In the early days of the IoT, it was thought that it was just a matter of hooking ‘things’ together. After all, how hard could it be? As companies started to develop IoT systems, they found out it wasn’t quite as easy as they imagined.

As more ‘things’ are added to a network, the amount of data being handled increases dramatically, putting pressure on the communications infrastructure. While 5G – when it appears – is likely to solve many of these issues, IoT developers realised that processing all information in ‘the cloud’ brought more problems. A ‘thing’ which sends data to the cloud is also waiting for the cloud to tell it what to do. Latency became another issue to address.

Then there is security; like communications, as more ‘things’ are attached to a network, the so called ‘attack surface’ increases. With the IoT potentially handling personal and other sensitive information, this has been a worry.

And there’s the question of power. One of the concepts underpinning the IoT is that ‘things’ can be put in all manner of places to collect data. But many of those ‘things’ will be difficult to get to. And the batteries which power these ‘things’ will need to last for many years, avoiding an almost impossible maintenance problem.

When you start putting these issues together, you are pushed towards a scenario in which the cloud becomes less important, with much more activity taking place at the edge. Instead of edge devices, we’re going to be seeing and hearing much more about edge processing.

NXP is one of the companies addressing the growing issues of edge processing. Until recently, it was expected that edge devices – the ‘things’ collecting information – would be designed around microcontrollers. But the issues relating to latency and, to some extent, security have brought a change of thinking; typical microcontrollers don’t have the necessary features, so why not use an applications processor? Well, some of the features of apps processors aren’t suited to edge processing. But if you can take the best of both worlds, you can create a device suited to the task. And NXP’s offering is what it says is a new class of device – the crossover processor.

“While the IoT tends to dominate at the moment, there are other things going on and the emergence of RISC-V could be disruptive in the embedded systems sector.”

Graham Pitcher

By focusing on edge processing, most of the data processing will be handled where the data is collected and communication with the cloud will only happen when necessary. Edge processing will therefore not only relieve the wireless network burden, but also improve response times and reduce data centre costs. As NXP notes, the edge is only going to become smarter and more aware.

That data will be collected by sensors and, as ams notes, sensors are becoming ubiquitous. And companies such as ams are developing new kinds of sensor with new levels of performance to support a range of innovative applications, including high end machine vision and autonomous vehicles. The sensors developed need to be easy to integrate, which means they need to come with software and applications support.

But even with more processing taking place at the edge, there will still be the need for those edge devices to communicate with the cloud. And industrial applications have specific needs, which are being addressed under the heading of the industrial IoT, or IIoT.

As Toshiba notes in its article, the sheer volume of nodes that need to communicate in future Industry 4.0 – or IIoT – architectures will make cabling each device increasingly impractical.

Those designing IIoT systems have a number of communications methods to choose from. Some may elect to use wired communications, but for those opting for wireless links, there is a wide selection. Many designers will be familiar with the benefits of Wi-Fi, including the ability to provide large bandwidth links. But Wi-Fi is a point to point technology; IIoT systems may well need something that enables the nodes in the network to ‘talk’ to each other – so called m:m communications. While ZigBee led the way in mesh communications, Bluetooth has caught up, with the SIG announcing earlier in 2017 that the popular protocol can also be used to create meshes. This, combined with the low energy (LE) variant, is set to make Bluetooth an attractive option for IIoT applications.

The IoT spans a wide range of computing requirements. We’ve already talked about processing at the edge, but there’s also the cloud to consider. As Intersil mentions in its article, the requirements range from mW to MW. Amongst the solutions to the challenge is digital power, maximising efficiency while delivering the levels of power required by the system.

Digital power also allows for ‘smarter’ systems; not only can they ‘talk’ to each other, the use of CMOS technology, where appropriate, allows for the integration of ‘digital smarts’. In the future, says Intersil, every bit of power will matter.

Knowledge on how to put IoT and other systems together will come from the distribution sector, says Mouser. If distributors are to stay relevant to the markets and customers they serve, it says, they must participate fully in the innovation process. This means not only providing the components designers need, but also becoming a knowledge centre, offering a wide range of useful content.

While the IoT tends to dominate at the moment, there are other things going on and the emergence of RISC-V could be disruptive in the embedded systems sector. AndMicrosemi, which is an enthusiastic promoter of the open source approach, believes RISC-V could open great opportunities in computer architectures.

Cadence, meanwhile, is building on its EDA heritage to support the emerging market for machine learning and artificial intelligence. It contends that systems won’t in the future be defined by chips, rather by the application in which the system will run. Applications-based system design, it contends, will open new opportunities for semiconductor and systems designers and for their supply chain partners. And key to this will be a system design enablement strategy, linking IC, package and PCB design that results in intelligent and elegant solutions.

One development which should be of interest to all designers is the latest addition to the UK’s Catapult network. The Compound Semiconductor Application Catapult is looking to enable the next wave of innovation by promoting the use of compound semiconductor. With more than 100 companies already in the compound semiconductor supply chain, the UK is in a strong position to move ahead quickly and the establishment of a technology cluster in South Wales, backed by significant Government funding, will support this.

One of the characteristics of the electronics industry is the speed with which it moves, which makes selecting the right technology a challenging task. We hope that Outlook 2018 will help you to make the right choices.