Two years after buying operating-system vendor Micrium, Silicon Labs has incorporated the software into the GeckoOS that the company will use to support its microcontroller push into internet of things (IoT) applications. It is just one example of how giving deeply embedded devices a network connection is changing not just development but the business models around it.
The Micrium-based software will not just accompany the discrete silicon but a growing family of modules that, in common with a number of other suppliers, Silicon Labs sees as vital in kickstarting development in IoT projects. In the run-up to major European trade show Electronica, the maker-market specialist Arduino has created its own IoT Cloud environment and launched a range of modular kits that combine Microchip microcontrollers with U-blox RF interfaces.
Fabio Violante, CEO of Arduino, says readymade modules and software tools that streamline development are vital: “IoT is a buzzword but it’s a complex problem to tackle. To do it all yourself you have to be an expert in many things: knowing the cloud APIs and knowing how to manufacture a board are not skills you find in a single person.”
Tyson Tuttle, CEO of Silicon Labs, argues offering libraries and an operating systems alongside its microcontrollers was a necessary step to address IoT applications. Because they need to handle real-time tasks, talk to remote IT servers and operate for years on a single battery charge, they have needs that are distinct from those of traditional embedded systems.
Suppliers such as Silicon Labs have coupled the software-programmable processor cores they use with increasingly smart peripherals that can function independently. This lets the processor sleep for longer periods, thus saving energy. As a result, the development tools need to coordinate software threads with commands to configure the peripherals to ensure data passes between the two cleanly. VWith IoT, the optimization has to happen differently: you are not going to have generic hardware in the platforms,” Tuttle says. “And they don’t want to write a lot of code. Users just to want to set things up: just have the wireless work and the cloud work.”
Juan Nogueira, senior director of connectivity centre of excellence at contract manufacturer Flex says the issues surrounding IoT development are often at a higher level in the business. He points to the confident predictions made by companies such as Cisco several years ago of tens of billions of IoT nodes being deployed by the start of the next decade. “We are nearing 2019. We don’t see the billions of devices. We don’t even see the millions. Why isn’t it taking off?” he asks, noting that many customers want to pursue IoT initiatives. “But customers are not sure about the comms technologies. Which sensing technologies? What data capture? How often to capture? Does this idea give them a return on investment?”
Violante says: ”The best way to understand the benefits of IoT is just to do it.”
Flex expects to pick up production business from IoT projects through its iENBL prototyping kit - making tuned versions of the hardware once the application has been trialled. Nogueira claims the contract manufacturer’s approach of putting as many sensors into the core product as possible would lead to more compact prototypes than sandwiches of carrier boards and shields typical of modular de facto standards such as Arduino and Beagleboard. However, Flex has not gone as far as Arduino and Flex in creating an overarching development environment, favoring instead the off-the-shelf open-source Eclipse tools.
“You can use Arduino and shields. These are good for initial tests to get a proof of concept. But how do you take it to the field? Are you going to take all these boards and batteries? You may have difficulty performing trials because the device is too big and bulky. And you may have to recode at the end if your production app isn’t based on the same architecture,” Nogueira argues.
Nogueira uses the example of a manufacturer of construction machinery that wanted to add location awareness to its products, using a combination of sensing modalities including pressure to detect height above ground and contact to identify tampering. He says the project started in April with an initial order of ten units for programming and benchtop trials. “In June they bought 150 units with customized color and branding,” he says. These were to be attached to live products in a limited trial.
The customer switched to a custom enclosure rather than the iENBL’s standard IP65 case in July and made the decision on which sensors to keep in a productionized version of the PCB. “In August, they changed the batteries they wanted to use but stayed with the core PCB design. In September we started producing the first product samples for test before going to mass production. We are expecting to scale to between three hundred and fifty-thousand and half a million in 2019-2020,” Nogueira claims.
Having built its reputation among makers and hobbyists with a combination of development tools and readymade modules built around the AVR and Arm architectures, Arduino has changed its licensing model to attract more corporate users.
Under the previous open-source license, a user would have to release any variant they designed to the public. “We have now created a dual-licensing model,” Violante says. The new license lets users, in exchange for a royalty, keep their designs as closed source.
Although Arduino is tackling low-volume production runs using its “Panini approach” of stacked modules and batteries, a new generation of small form-factor boards allow greater hardware customization. The Mkr board is based on the combination of an Arm Cortex-M processor and one of Intel’s Cyclone 10 field-programmable gate array (FPGA).
In a number of applications, module size may not be an important factor. Violante points to two users where IoT modules have been added to heavy equipment, with one of them enabling a novel business model. Fluid Intelligence, he says, now offers “connected oil” to its stone-crushing machine customers. The IoT hardware is an Arduino module fitted with RS485 interfaces that communicate with specialist sensors. “The data goes to their cloud where they apply their algorithms. They offer subscriptions for oil management that is used to reduce the downtime of the equipment. It extends the life of the oil by one or two orders of magnitude.”
Another user supplies agricultural tractor components and is now retrofitting existing vehicles with tracking modules. “Buying a digitally connected tractor can be €100,000,” Violante says. This option is somewhat cheaper for users. The module uses the LoRA low-power wide-area wireless network (LPWAN) network for relaying location data to the cloud and a CAN shield communicates with the tractor electronics.
“They track everything. They make sure the right driver is on the right tractor and that the right accessory is connected. If the pesticide is going to the wrong place, they can stop the tractor from the cloud,” Violante claims.
Naturally, going from a position of where tractors can go where they are driven to being given a degree of remote override changes the nature of how such large system are used – and demonstrate the requirement for full prototypes and field trials to work out the niggles not just in the software but the business models.