The Limits Of Energy Harvesting

Energy harvesting, once considered an inexpensive alternative to low-power design and a way of achieving nearly unlimited power in mobile devices, has settled down to more modest expectations.

This approach to generating energy through a variety of means—from solar to motion to ambient RF and even pH differences between soil and trees—has been proven to work. The problem is that it doesn’t generate sufficient energy quickly enough for most applications. And like most analog circuits, it doesn’t shrink. Energy-harvesting modules tend to be bulky, adding area and cost to designs.

Fig. 1: Thermal energy harvesting device. Source: Delta

“If you’re not burning a lot of energy, there is no need for energy harvesting,” Gert Jørgensen, senior vice president of sales and marketing for Delta’s ASIC Division. “And if you do burn a lot of energy, then energy harvesting will not work. Where we see this working is in the 10 milliwatt to 100 millwatt range, but a battery is still 20 times cheaper. Energy harvesting technology is big. You have the harvesting element, power management and storage. If you compare that to a battery, the battery is smaller.”

When the first smart watches hit the market in 2014, most of those devices were based on 28nm or 40nm logic. Consumers complained that some of these watches didn’t last more than a day before needing to be recharged. Advanced power management capabilities could be built into these devices—dynamic voltage frequency scaling, multiple voltage rails, dark silicon. But none of those would be sufficient to make batteries last more than a day if these watches were constantly connected and being used for processing-intensive applications such as real-time health monitoring.

The idea at the time was to utilize motion or ambient RF signals to produce energy, similar to the way a self-winding watch works. So far, that has not panned out.

“We’ve heard about voltage generation from electromagnetic radiation and exotic approaches based on gallium arsenide and gallium nitride,” said Vic Kulkarni, vice president and chief strategist in the office of the CTO at ANSYS. “There has been experimentation on looking at the impurities in gold where you can make electrons jump, and there are new materials that are lightweight composites that generate energy. But right now, this is happening only in a very small way commercially.”

That seems to be the general consensus, barring a major breakthrough in materials engineering.

“Fundamentally, energy harvesting is a technology with a limited use envelope,” said Jeff Miller, product marketing manager at Mentor, a Siemens Business. “It has to be an edge sensor. There needs to be a source of energy, whether that’s heat, mechanical or solar. And it has to consume more power than can be provided by a battery. A battery is a reliable option, and batteries can last a long time. If you think about a vehicle tire pressure monitor, that’s a tough place to change a battery, but you still can get 10 years of battery life.”

Energy harvesting options
The most successful types of energy harvesting today fall into two camps—solar and motion-driven. Solar uses photovoltaic cells to convert light into electricity, but efficiency of today’s photovoltaic cells is still in the low- to mid-20% range. In addition, solar takes up a lot of space, and energy-intensive applications such as a smartphone or an electric vehicle use far too much energy to make solar a viable option. But small-scale photovoltaics still play an important role in calculators, watches and small devices.

“Solar is still the ultimate in energy harvesting,” said Delta’s Jørgensen. “It’s mature, it produces direct current, and it creates a reasonable amount of energy.”

Motion-driven energy harvesting is the second major approach in use today, and it’s a difficult one to deploy on a chip or inside a small device.

“On-chip is very difficult because of space,” said Peter Harrop, chairman of IDTechEx. “It isn’t going anywhere fast if you’re looking at energy harvesting with comb-like, capacitor-like structures or vibration harvesting. You tend to meet the wrong vibration with piezoelectrics. The vibration you can harvest is very narrow, and it drops off immediately. It’s the same with acoustic frequencies. You tend not to meet the frequencies that you need. Other than self-powered sensors, this is very limited. So on the chip, it’s not looking too good. And in the case of using the chip inside an Internet of Things node, the dream of selling billions of Internet of Things sensors every year isn’t happening. It’s millions, not billions. And if you’re going to put one on every tree in California to spot which way the wind is blowing to make sure you’re not downwind of a fire, you can’t change the batteries so you have to have energy harvesting that works, but that is not a solved problem.”

Where energy harvesting really shines is on a much larger scale—solar, wind and wave motion. At the chip or small device level, use of this technology is much more limited. Up-and-coming new technologies include piezoelectrics, thermoelectrics, photovoltaics, triboelectrics and dielectric elastomers, most of which have been developed over two decades of research into nanogenerators at research centers such as the Georgia Institute of Technology.

Piezoelectric power sources can be built into diagnostic skin patches and implants for medical devices, layered on as paint to power inset LEDs, added as layers in integrated circuits and in MEMS chips. They also can play a dual role, powering and providing processing to modulate light in a way that enhances the performance of photocells and LEDs, or integrated with photonics for high-bandwidth data center data-access applications.

Piezoelectric power sources are better known, but Georgia Tech researchers have had more success with triboelectric energy nanogenerators (TENGS), which use friction and electrostatic induction to generate small but usable bursts of power. The result is a method able to deliver a steady flow of DC power at 1.044mW using onboard battery storage and a collection process able to function at up to 300° Celsius, using a dual-capacitor method that is up to 60% efficient and uses on-board storage, according to a 2015 paper in Nature Communications, and a 2018 paper in the journal Advanced Materials, which confirmed static charges come from electrons, not ions.

TENGS offer a range of power levels, often at a lower cost than batteries or alternative methods of harvesting energy. It also uses non-poisonous materials in fabrication, and design simplicity that should make large-scale manufacturing simple.

Energy harvesting techniques also look like a natural fit for small sensors and IoT devices that may be used in places where the batteries can’t be replaced easily, according to Linley Group analyst Mike Demler.

But the consistency, cost and lifespan of devices running on ordinary lithium-ion coin batteries – along with the power-saving designs created and refined by chipmakers for decades—leaves most of the traditional semiconductor market with no good functional or economic reason to try to switch, Demler said. And while that could change in wearables and medical devices with specialized requirements, but there has been little mainstream demand so far, he said.

Even in devices that wouldn’t benefit much from the ability to reclaim power from the environment around them, there is a growing need for far more refined power management and, more importantly, control over the electromagnetic interference it causes.

A mobile phone likely to be sold in Europe and Asia, for example, would have to have RF circuits supporting as many as 30 different frequency configurations for different carriers — circuits that are all part of the same laminate, but often in different layers. It also would require filters and amplifiers to maximize gain and minimize EM interference in a design that quickly becomes “mind-bogglingly complex,” according to  Larry Williams, director of product management at ANSYS. “Impedance affects the behavior of the circuit, but you can’t just use a behavioral simulation of that for everything because the drive loads differ depending on the device. And they change as the temperature changes. So you have to pay attention to the way the IC die is attached to an SoC or circuit board and how they all react to changing temperatures.”

Power supplies are good and reliable enough that the need for power harvesting is relatively rare, but lack of attention to power architecture can ruin that even in low-power IoT devices, according to Jerry Zhao, product management director in the Digital and Signoff Group at Cadence.

“The methodology to design mixed analog/digital signals is critical,” Zhao said. “There are devices that primarily deal with the analog world — harvesting analog signals and converting them to digital efficiently is a challenge on the design side. It’s always a challenge. The analog is always a big chunk of the silicon and the model has to make sure the analog side will understand the power delivery. The flow is critical.”

It’s not just low-power IoT devices that make fine-tuning power management a difficult, however.

“The challenge is that the chip is getting bigger, with more and more devices on it and the power supply is lower,” Zhao said.  “Right now you have 0.6 or 0.8 volts to drive billions of devices on a single die — with what could be dozens of power domains on a single die, not just one or two. Analog power domains, memory power domains, those you can switch on and off to be more efficient for certain activities. The wakeup process takes time and drains a lot of current, and sometimes the current is so high it destroys your power grid. That is what needs to be analyzed at our power level. So we’re not looking at how to harvest the power from waves or wind or other sources in the environment. We focus on a micro level to make sure you get the performance required for all those devices without wasting power.”

Conclusion
Energy harvesting will continue to evolve at the chip and small device level on a number of fronts, but whether it will be considered reliable enough and inexpensive enough to compete with improvements in battery technology remains to be seen. Nevertheless, some of the ideas that are developed on a small scale can be applied on a much broader scale.

This remains a nascent but intriguing area for research. But what will achieve commercial success and when is as much a mystery today as it was five years ago.

Related Articles
Make Your Own Energy
Efficient use of power and energy in electric vehicles and smart buildings will require innovative thinking.
Where Is Energy Harvesting?
Sensors that have to be smart enough for networking and security have power conservation all set, following a decade of focused effort by chipmakers.