Realizing the full potential of the Internet of Things will require a cohesive hardware and software design approach.
With the rise of smart cities, cars and houses, an enhanced connectivity infrastructure bolstered by an increasingly connected culture, the Internet of Things (IoT) represents an exciting opportunity for semiconductor industry players.
As such, market researchers at IDC expect the installed base of the Internet of Things will be approximately 212 billion “things” globally by the end of 2020 including 30.1 billion installed “connected (autonomous) things” in 2020 largely driven by intelligent systems that will be installed and collecting data — across both consumer and enterprise applications. The research firm expects IoT technology and services spending to generate global revenues of $8.9 trillion by 2020.
Bryan Lawrence, segment marketing manager at ARM, noted that the IoT space spans sensor to server and existing software and hardware paradigms, and these areas already are being disrupted.
Key enablers to this will be tiny, low-cost sensors with connectivity that also will require secure, standardized Internet access with authentication and trust that will drive adoption of this technology, he said. “The way that each end node provides its data will also need the industry agreement so we have universal data model semantics that allows interoperability.”
Eran Briman, VP of marketing at CEVA pointed out the obvious: IoT has become a major buzzword. He noted its wide diversity – from wearables to health and smart home, and smart cities to automotive, among others. Specifically, “wearables is definitely an area that mandates ultra-low power design because there is an ‘always on’ requirement and a need for contextual awareness. These devices must be always listening, watching and sensing. And the battery capacity and the expected time between charges means power efficiency is critical. Requirements for power consumption are an order of magnitude lower compared to mobile processors.”
For these applications, there is a clear need to use the most appropriate silicon resources (i.e. engines, processors, etc.) in terms of power efficiency, Briman said. “One really needs to fine tune the system architecture for the requirements at hand. Given the nature of some of the processing (e.g. voice activation, face triggering), ultra-low power DSPs specifically architected to deal with these applications are the optimal choice.”
Navraj Nandra, senior director of marketing for DesignWare analog and mixed signal IP at Synopsys, also noted the market diversity that seems to be focusing on wearables, smart meters, and machine-to-machine applications.
In terms of the particular blocks that get used, he said, design teams are looking for analog front ends that connect into digital processing units, where the analog front end – such as an A-to-D converter — takes in a signal from the environment (voice, lights or any kind of analog signal), converts it to the digital domain, gets processed by the microprocessor, then gets sent out somewhere else by a D-to-A converter converts it back into the analog domain. “What you’re looking at from the analog perspective is something with a fairly low speed and reasonable precision that has to consume very little power. Its switching frequency is typically in the few hundreds of kilohertz.”
Technologies that play key enabling roles for IoT from an IP perspective include a user interface (touch, voice, gesture); low power or soft power; wireless connectivity (2G, 3G, Bluetooth LE, Zigbee); security; low power processor/controller; audio codecs; sensor subsystems; non-volatile memory that could be used for storage or security; A-to-D converters; USB; MIPI storage, and others.
ARM’s Lawrence also observed that many of the end nodes in the IoT will be in objects that provide small amounts of data infrequently. As a result, many of the devices will have to be in extremely low power hibernation states most of the time. Because of this they will rely on energy harvesting, solar, heat, vibrations to keep them alive, which will be measured in nano amps (nA). He noted that ARM’s silicon partners already offer microcontroller devices that work in this manner with very sophisticated sleep and hibernation states from which the device can wake up reasonably quickly, process a sensor measurement and transmit the data before returning to a sleep state or hibernate. There also are ultra-low power radio designs that can run for 10 years from a button cell battery.
Extremely low power versions of processor cores are being rolled out to support these opportunities. Synopsys offers an ultra LP version of its ARC processor. ARM offers its Artisan 90ULL platform for TSMC process technology, well suited for implementing IoT designs due to its combination of power-optimized memories and logic libraries. Even Intel has jumped into this market with its Quark ULP processor. Further, STMicroelectronics announced this week a collection of analog and mixed-signal devices aimed at wearable applications.
However, to truly enable the IoT, it comes down to the software development community. Here, Lawrence said, an easy, open environment is needed to bring in many more developers who are currently designing apps and webpages.
Andrew Caples, senior product manager for Mentor Graphics’ Nucleus product line agreed. “We see this huge push by our semiconductor partners who are spending lots of R&D dollars and lots of R&D cycles putting low-power features into their processors and controllers.” Some processors have more than 10 low power states and contemporary processors today have dynamic voltage and frequency scaling (DVFS), various sleep modes, low-power modes and hibernate modes being architected in the silicon.
“What’s more, you’ll see from their positioning their performance relative to power consumption is always a big parameter that they want to highlight to talk about their efficiencies,” Caples said. “You see techniques with superscalar processors, for example, that are using virtual cores as a vehicle to get more performance out of the silicon without adding any additional power consumption to it. What we see on the software side is the need for power-efficient design for various reasons. One is to expand the battery life, especially when you talk about IoT, and many of these will be on dispersed networks, many will be portable or battery operated devices and you want to extend the longevity. Additionally, for non-battery-operated devices, green energy is here. It’s real. Devices need to be more power efficient, and that starts not just with the silicon you select, but also with the software and how you design it and the framework that’s underneath.”
And while there is a lot of effort in the silicon for low-power features and a desire from the semiconductor companies to develop power efficient systems, from a software perspective there is a void in the market for a framework that allows software developers to write power-efficient code, he pointed out.
“If you look at the amount of code that sits between the hardware and the application developer — and when you are talking about device drivers, board support packages, operating systems, you have to include the middleware — typically all of that is ported and delivered to the application engineer. The application engineer is asked to write code that is power efficient, but they are a long way removed from manipulating or taking advantage of the low-power features of the silicon. They’ve got all of this middleware and operating system and drivers in between, and it is very difficult at that stage to start writing code that is going to switch an operating point in the system. The software can only take advantage of the features that the hardware provides. The hardware sets the framework for what power efficiencies can be achieved, and then if you have a framework that allows the software to take advantage of those hardware features then you can maximize what the hardware offers.”
Looking at this from the industry level, the IoT will reach its full potential when the software and hardware design ecosystem understands these issues and works together to solve them.