This means that the one-size fits all embedded processor model that characterized the portable device market breaks down when applied to the Internet of things (IoT).
The example of the connected home provides evidence of this contention. According to research estimates gathered from ABI Research, TechNavio, Pike Research, and BI Intelligence; connected-home devices will take the form of
· Energy devices: smart thermostats, smart lighting,
· Safety/security: monitor, cameras, alarm, and
· Smart appliances: washers, dryers, refrigerators
Each of these categories as well as individual products within the categories will need unique processing requirement to meet the cost, power, and performance provisions of their individual applications. Besides the smart home, health care, automotive and industrial will also demand individually unique embedded computing solutions.
One design that illustrates the unique computing requirement of an IoT device is the electronic shelf label (ESL). Retailers use the ESL to display product information and price on their shelves. A liquid crystal display or other technology, such as electronic paper attached to the front edge of a retail shelf, displays the product information.
The product pricing and other information displayed on an ESL is automatically updated whenever changes occur. An Andes customer uses a 60-MHz AndesCore N801 core with 128KB of eFlash to control an ESL device that communicates over a sub-GHz frequency band with fast 2-way transmission (see Figure 1 below). The application uses a high-resolution e-paper display and requires an ultra-low power N801 to help run for 5 years.
Figure 1. M2 Communications M2C8001 SoC in an Electronic Shelf Label Application
Embedded computer requirements for IoT devices, such as the ESL, include security, communications, sensing and control, and power management. Security is essential to prevent network attacks as well as physical attacks and to protect against software/firmware theft. Communications requires megabytes of processing to handle proprietary or standard protocols such as RFID, 802.15.4, Bluetooth Smart, Bluetooth 4.1, WiFi 802.11 ah, and LTE Cat-0. Processing a variety of sensor data coming from the control interface requires versatile DSP capabilities. And power management is essential to enable months to years of operation on small batteries or harvested energy, thus demanding processors with efficient power management for long sleep cycles, fast power-up/power-down, and the ability to operate at varying clock frequencies—to sip energy rather than full-on or full-off.
To serve this range of requirements, embedded processors need to provide general-purpose computing with high performance-efficiency and flexible power consumption. They must offer DSP computing capabilities such as SIMD (single instruction multiple data) processing to efficiently handle the wide range of sensors operating with these IoT devices. Finally, these embedded processors need to be extensible to allow application specific acceleration and added security. This is the architecture Andes created in its new generation of embedded processors developed in 2006 and enhanced over time since then to adapt to IoT application needs.
For example, the Andes Technology secure microprocessor unit, shown in Figure 2, and S8 CPU core together provide security from hacking. The secure microprocessor unit (MPU) provides functions not found on competitive 32-bit CPU core offerings: data and address scrambling and differential power analysis protection. The first provides defense against hacks that target the interface between CPU and memory. The second guards against hacking the program by observing the power use signature of the CPU.
While having an elegant architecture addresses the problem on paper, the real solution results from being able to realize the complete solution in a final IoT device. As a CPU IP supplier, Andes provides a complete pre-integrated low-power SoC IP development platform to realize compact code size, low power, and secure debug.
This platform comprises a full complement of peripherals surrounding the CPU core, a software development environment with real-time operating system, device drivers, and the software stacks that typical IoT applications require. This is all supplemented with FPGA-based and silicon-based development boards to prototype the IoT applications before producing final silicon and/or final devices.
Over the remainder of this decade, SoC designs to implement Internet of Things devices will proliferate. These IoT designs will differ greatly from those going into mobile devices of today. IoT chips will require an order of magnitude less power consumption, while still demanding high performance and a minimal silicon footprint. Only CPU cores based on next generation architectural features such as those Andes Technology pioneered can hope to meet these stringent requirements.
About the author:
Charlie Su is CTO and Senior Vice President of R&D at Andes Technology Corporation – www.andestech.com