As an industry, we’re still in the early days of wearable technology – figuring out what works (and what doesn’t) and continuing to define and refine this ever-growing market. As David Maidment of ARM says, wearable devices make up such a large and diverse category.
Their architecture and development requirements speak to that diversity – a smartwatch does not have the same demands as a fitness tracker; a wearable camera does not fit in the mold of a health monitor. But, as varied as this category can be, there are commonalities and shared lessons.
For basic devices – things like small fitness trackers or limited capability smartwatches – the key is on-chip memory and ultra-low-power processor cores. These provide enough processing power to gather data from always-on sensors while allowing for a simple Real Time Operating System (RTOS). These low powered chips can run at 20- to 150 MHz and can provide months of battery life – making this type of architecture perfect for an always-on device designed to become part of a user’s day-to-day routine. This architecture is tried and true, offering the power efficiency needed for these specialised devices.
As we get into mid-range devices – such as smartwatches with more rich operating systems and color displays – balancing processing power with battery life becomes even more important. Here we start to see a need for processors running between 200- and 500 MHz with an RTOS or even a full OS based on Linux or Android. This setup allows for a constantly connected device that can offer a rich user experience without sacrificing battery life. To achieve this, we look for chips that feature sleep modes designed to conserve power when the user is not directly interacting with the device.
And then there are the high-end devices – more robust smartwatches, for instance. These devices may require up to a dual-core cluster multiprocessing application processor that allows for scalable performance. In this case, the best course of action would be to split the load – a powerful processor for when the device is in direct use, and a power-efficient one to provide always-on processing support. This gives you the best of both worlds – the processing power of a high end core without sacrificing energy efficiency. This allows for processing power between 500- and 800 MHz, designed for Android and Android Wear devices, while offering low active power and low power modes, along with low-power DDR memory.
In each of these cases, we’ve achieved a device that offers that much sought after balance of function and convenience, a device that does what the user needs while being as convenient and transparent as possible. Across this broad spectrum, we see common challenges – balancing the processing needs of a modern, connected device with the things that make a wearable device appealing to a user – wearability, long-battery life and a rich user experience. While the industry’s experience with smartphones and other mobile devices has been invaluable in tackling these wearable challenges, these devices offer a far stricter sense of design constraints than other mobile technologies. To build a wearable device, we must optimise the SoC, which include the use of smaller data memories, slower clock speeds and the most power efficient silicon technology processes. For instance, you can use smaller memory caches that save on die area and power – it’s possible to halve the L1 cache size from 32K to 16K but only see a 10 percent impact on performance. These are the kinds of efficiencies that are needed to truly address wearable challenges.
At its core, making a wearable device is an exercise in balancing performance and power consumption. With the broad range of form factors and use cases for wearable devices, there’s no one-size-fits-all solution. But, we’ve learned a lot from the smartphone revolution and we’re continuing to build and innovate on that progress each day. As processors get faster, smaller and more energy conscious, wearables become even more exciting. We’re excited to see what they will look like in even a few short years. As an industry, it’s important to recognise that, at their core, wearable devices contain lessons to be learned and common standards to build from – and sharing these lessons will be key to making sure that innovation in the wearable space is given its very best chance to succeed.
The author, David Maidment is Mobile Segment Marketing Manager at ARM where he looks at the application of ARM technology in the mobile space areas including wearable devices and mobile platforms.