Building Billions Of Batteryless Devices

Later this month, Arm will celebrate its 30 year anniversary and the engineering milestones that have resulted in more than 180 billion Arm-based chips being shipped in everything from sensors to smartphones to the world’s fastest supercomputer.

In each of these cases, much of Arm’s success has been in our dedication to delivering the highest performance per watt. But while Arm may have written the book on low power compute, it’s in the far more constrained world of ultra-low energy devices that my team and I have been actively working for some time.

Earlier this year, we developed a prototype sub-threshold MCU as part of the DARPA N-ZERO program. In remotely deployed sensors, an MCU needs very low sleep power to maximize battery lifetime, deterministic real-time response to capture rare sensor events and energy-efficient operation—with enough compute and memory to run useful workloads. In standby, the M0N0 processor draws a phenomenally low 10 nanowatts, rising to between 20 to 60 microwatts/MHz depending on the active application. As Simon Segars put it in his Arm DevSummit blog earlier this month, a single 30mAHr button cell battery could theoretically power the M0N0 SoC for 340 years.

In reality, that button cell would lose its charge due to leakage far more quickly than that. Batteries are finite—and so are the reactive metals they’re based on such as lithium, the mining of which is set to triple by 2025 in response to the growing electric vehicle (EV) market.

But while prices may have dropped recently, production can still be prohibitively expensive, making or breaking a device’s design and functionality. That’s not to mention the potential environmental impact of greatly increased battery manufacture: The pervasiveness of billions of devices can’t come at the expense of the planet.

The prospect of a better way to power future IoT devices led my team and me to begin investigating alternative power sources. Could it be possible, we wondered, to create programmable devices which are not only wireless but batteryless too?

The idea of batteryless Arm devices was a challenge too tempting to pass up. And it resulted in what we’re calling Project Triffid.

Batteryless Arm devices: A whole new level of low

Batteryless devices such as solar Bluetooth low energy (BLE) beacons are available on the market today for around $20 a unit, but for many potential applications that’s simply too expensive. We believe to truly unleash the largely untapped potential of pervasive computing we need to build computers that cost only a few cents to produce. By significantly reducing costs and extending device lifespans we are lowering the barrier to workable application areas for secure, smart sensors—bringing the benefits of advanced, low power computing to more applications than we can currently dream up.

Which brings me to one of the cheapest and most pervasive devices, and where Project Triffid began: RFID tags.

Passive RFID tags exist in their billions around the world. They’re in our credit cards, our cars, even our pets. And of course, they’re used in a wide range of applications throughout industry. Passive RFID doesn’t contain a battery: it harvests electromagnetic energy from the device used to read it. While it can receive and store information, it’s a ‘dumb’ device—it can’t do anything with the information other than offer it up when asked.

Computational RFID: A huge step forward in sustainable devices

Passive RFID tags exist in their billions in our credit cards, our cars, even our pets.

That’s where Arm’s Project Triffid comes in. Theoretically, it’s possible to embed a very low-power or even batteryless MCU capable of simultaneously managing both wireless power and wireless data, inside an RFID tag. This would enable the tag to not only capture and store information but act upon it—whether compressing data, analyzing trends or detecting anomalies.

Computational RFID (CRFID) applied in this way could be a giant step towards the enablement of billions of sustainable Arm devices.

Intermittently powered computers, which survive off energy harvested from their environment, could monitor objects in remote places, without maintenance, for decades. CRFID systems use scavenged radio frequency energy emitted by ceiling- or drone-mounted readers (for example), and can be suitable for deployment in environments where other harvesting systems may fail. Developing a system like this is far from straightforward—especially within our strict constraints. We want the system to have a 10 meter range from the reader in order to minimize overheads and maximize uptime, but this likely leads to a measly 1uW power budget.

We’ve built battery-based systems that could run at this level, but we knew the battery was stable and didn’t have to deal with regular and random power failures that may happen hundreds of times per second. In order to mitigate this intermittency in batteryless Arm devices such as Project Triffid, we need to develop circuits to detect it, software that can react and adapt appropriately and non-volatile memory to maintain the work-in-progress. All without blowing that 1uW budget!

We know there are research teams out there who are making good progress in this area. For example, we’re collaborating with the University of Southampton on antennas and simulators, and they have been working on intermittent computing for years. As well as publishing some impressive work, they are also making moves to broaden the application of transient computing by adding Arm Mbed support to their technology. Whilst this work is all moving in the right direction, as far as I’m aware, there’s not yet a complete system that brings all the necessary components together to make economical batteryless Arm devices a reality.

Extensive potential use cases

We strongly believe that Project Triffid technology has the potential to positively impact not only our day-to-day lives, but also the world around us. Imagine being able to place a truly smart tag anywhere, with no constraints, which lasts as long as you need it to—potentially forever.

The potential use cases for Project Triffid are extensive, particularly in areas such as perishable goods: a Project Triffid tag might employ light or temperature sensors in order to report whether goods have been stored incorrectly. Combined with the ‘e-nose’ technology we’ve been exploring as part of our research into plastic chips, Project Triffid could even be used in retail food packaging to warn if food is unsafe to consume.

While that application would end once the food was consumed, others extend to the product’s entire life cycle: take a smart sneaker, for example. An embedded CRFID chip could enhance the supply chain, getting the right pair of shoes efficiently to the right customer, with minimal friction. It could provide the user with data about their gait or activity levels without them having to remove and recharge a device.

And at the end of the sneakers’ life, it could aid with returning materials to the manufacturer and ensuring optimal reusability of each component, improving sustainability and progressing towards a circular economy. The same batteryless Arm devices could send privacy-preserving usage data back to the manufacturer, with insights that could improve future designs—all with little to no human interaction. Through Project Triffid, previously ‘dumb’ objects could be transformed into useful tools for understanding and optimizing products and processes.

There are a huge number of potential applications for batteryless Arm devices, highly deployable embedded devices. But what we really need is for the industry—the people who are out there dealing with real-world challenges—to tell us how they might be able to usefully apply this technology to help us create something that could really make a difference.

What’s next for Project Triffid?

Our team has worked hard over the past few months to reach our first tape-out milestone, and we’re currently testing our first set of 28nm chips. To take the project to the next level, we plan to tape out the next iteration of chips at 22nm, enabling the incorporation of MRAM. We’ll be publishing the results to boost ecosystem development and help make affordable batteryless Arm devices a reality.