These devices will be connected via low-power wireless networks, which will also require intelligence for management and functionality, via internet gateways and up into the cloud, where a mass of collected data can be accessed by mobile devices or analysed for IoT use cases.
For these IoT applications to be successful, crucial characteristics of end-point devices are very low cost – cents rather than dollars – and minimal energy consumption: devices will be powered by coin-cell batteries or scavenge energy from their environment and will need to last months or years rather than hours, days or weeks. A further typical characteristic of end-point devices is that they will usually be in sleep mode with infrequent communication.
Further application requirements include security, which is absolutely crucial in the IoT, and it must start within the end-point device and from there through the network and into the cloud. Furthermore, application software deployment on devices needs to be easy to access and control for maintenance and updating system or device parameters. Systems will need to be highly scalable and operate efficiently whether there are just a few or thousands of devices in the network.
Challenge – three engineers in three months
As in any electronics market – given sufficient demand for volume production – the advantages delivered with the integration of functionality on system-on-chip ICs (SoCs) include BOM-cost savings, robustness, size and power efficiency. Unlike leading-edge mobile consumer devices such as smartphones, however, IoT products and especially end-point devices such as sensor-based environment-monitoring systems will require a completely different performance/power profile, with low energy consumption typically being the overriding consideration.
In a proof-of-concept project, ARM set itself a challenge in 2015 to demonstrate that the physical implementation of an SoC for an IoT end-point device with strict low-power and cost constraints is easily attainable for small design groups from companies of almost any size, including start-ups as well as large OEMs/ODMs. Furthermore, the goal was to develop a platform and prototype a test chip that would enable their design teams to rapidly build differentiated solutions by integrating varying combinations of ARM IP with their own IP.
The project should also demonstrate the potential of ARM’s platform for IoT to minimize risk, while also reducing cost and accelerating time-to-market with only minimal engineering resources. One further goal was for ARM to gain a better understanding of the challenges of design and IP implementation for IoT devices – for example, it was an integration-first in a design for the company’s new Bluetooth radio IP together with embedded Flash.
The specific task was to rapidly implement this prototype silicon demonstrator platform – called Beetle – with only three engineers and in less than three months. Finalised and taped out in Q3 2015, the resulting platform integrates ARM IP on a single piece of silicon and includes an IoT subsystem with Cortex-M microprocessor, Bluetooth ‘Smart’ Low-Energy (BLE) radio, plus embedded Flash memory – see figure 1.
Fig. 1: The Beetle test chip die (right) and its different blocks (left).
The Beetle being a test chip, minimal optimisations were performed for area or power management. But even without significant effort, initial benchmarking and analysis were in line with similar ARM Cortex-M3 based devices. The ARM Cordio radio operates in the sub 1 Volt region as per design, and therefore allows extremely low power communication.
This test chip was built using the ARM Artisan physical IP platform specifically tailored for IoT applications. The design is also fully compliant with ARM’s mbed IoT platform to enable rapid development and prototyping.