Where Is Energy Harvesting?

With power management a top priority in sensor networks, why is energy harvesting—a proven technology with diverse energy sources—conspicuously absent from sensor designs that are the foundation of the Internet of Things?

Energy harvesting always has been a promising answer to the limits of battery power. The idea that a device can run for much longer periods of time between battery charges with continuous or even sporadic energy infusions has been a topic of discussion for years, particularly as demand for new features are added into portable electronics and the demand for more compute power continues to rise. But actual adoption of this technology has been much slower than initial projections suggested.

Even several years ago, as the IoT was still being defined, there was strong consensus that being able to constantly generate enough energy through various types of energy harvesting would replace the need for more sophisticated power management both on- and off-chip. Nevertheless, the cost of the energy harvesting technology itself, along with the low amount of energy that can be consistently harvested from a spectrum of possible energy sources, have largely kept this technology out of devices where it appears to be an obvious fit.

In contrast, there has been significant progress in energy management in chips. Case in point: A new smart phone has significantly more functionality than what was available in smart phones sold several years ago, with equal or better battery life. Most chipmakers have opted for architectural optimizations rather than focusing on energy harvesting, scaling to the next process node, using low-energy Bluetooth or Zigbee rather than cellular networks, and relying on a host of refinements and modifications in designs.

“People are still grappling with power and harvesting,” said Vic Kulkarni, vice president and chief strategist for the Semiconductor Business Unit at ANSYS. “Customers used to ask about power reduction. Now it’s energy management and thermal management.”

It’s not that research in energy harvesting is slowing. New papers continue to show up on energy harvesting from photovoltaic, piezoelectric, biochemical, thermal, and mechanical vibration sources, especially for wearables such as watches that power themselves with motion. But its adoption in real products has been slow, and that doesn’t seem to be changing.

“I haven’t seen much commercial adoption,” said Ron Lowman, product marketing manager of the Solutions Group at Synopsys. “There is incremental development in a lot of areas, but people are looking for high-volume runners. I haven’t seen a lot of push to develop them. There’s always a higher cost to implement these kinds of technologies up front, and people are risk-averse.”

Using energy harvesting or wireless power transmission adds uncertainty about consistency of power, as well, while existing approaches will provide at least some power under all conditions.

“If the energy supply isn’t really under your control or is subject to what the environment provides you, you have to take avery conservative approach to power management of that device,” said Jeff Miller, manager of product marketing and strategy at Mentor, a Siemens Business.

A decade ago it was far more common to see designs that incorporated energy harvesting in devices — such as smart environmental sensors — that monitored their locations around the clock and maintained wireless connections to send regular updates. These designs required battery changes only every 5 or 10 years, explained Phil Solis, IDC’s research director for Enabling Technologies.

“A lot of effort has gone into designing chipsets to consume so little power that they don’t need extra assists to make the battery last 10 years or sometimes longer, which is the goal in IoT,” Solis said. “These chipsets are engineered to be really energy efficient anyway, and they sleep almost all the time. They keep as little powered as possible, wake up, send an update and go right back to sleep, so they get a long battery life. They’re probably more reliable without adding another element.”

Growth in some sectors
Still, don’t count energy harvesting out. Europe is enthusiastic about the concept. In fact, the International Electrotechnical Commission (IEC) ratified the EnOcean specification in 2012 as an ISO/IEC standard for wireless applications with ultra-low power consumption, including specifications for energy harvesting as well as energy transmission.

The market for energy harvesting technology also is expected to grow from $311.2 million in 2016 to $645.8 billion by 2023, according to MarketsAndMarkets.

Energy harvesting is becoming more common in more complex devices, but the cost of the technology today is limiting its commercial appeal. Walmart, for example, is calling for sensors that cost less than $1. The retail giant uses the sensors to help maintain food quality and make its global supply chain more efficient, according to a report by the Harvard Business School.

Equal pressure comes from the opposite direction, as well, with sensors able to communicate using Bluetooth, , Zigbee, WI-Fi, as well as over 2G, 3G or 4G cellular networks. And they still need to have enough processing power to make those networks secure enough.

“IoT devices were supposed to be simple,” said ANSYS’ Kulkarni. “Now we find we’re in a multi-domain analytical situation where we have to do energy and thermal analysis about the chip versus the package versus the system and antennas. Everything is connected and has to be taken into account in terms of size and weight and energy.”

This has put enormous pressure on chipmakers to develop devices that are semi-customized, but with price tags that are comparable to mass-produced chips. Investing in new technologies in that context is economically unappealing.

“IoT is all over the place—there are a thousand different applications,” said Synopsys’ Lowman. “There are definitely sensors, but they’re networks of sensors that have to communicate across a network. You see the advent of voice recognition and personal assistants. People talk about mobile AR/VR. But those are really just mobile chipsets being re-purposed—security cameras, cams in general. There are often size or weight restrictions, and a requirement that we keep the customer in a lower-cost chipset. It’s a challenge.”

Security
Meanwhile, demand is increasing for security, which needs compute power and storage to support authentication and encryption, and two-way communication with enough diagnostic or remote-control capability to know if a specific node has been compromised.

“There are billions of nodes,” Kulkarni said. “If some become hostile, you have to see if someone has taken them over and you may have to decommission a series of nodes globally, so security becomes important in node intelligence, not an afterthought.”

Even physical security is an issue in some cases, Kulkarni said, because attackers can examine the heat signature of a chipset to identify potential points of attack. Some customers have begun asking for increases in power use in some areas of the chip to even-out hotspots in the heat map and make weak spots harder to identify, he said.

Any of those additional features, or even the addition of antennas to support the requisite number of networking protocols, affects the power profile enough to make the device unacceptable to a customer even when requested features are all in place, Kulkarni noted.

Reducing power
The cleanest way to deliver more power with less energy is through process node reductions, from 28nm to 16/14nm, and even to 10/7nm. That has limits, however, because moving signals across skinny wires from one side of a large chip to another raises issues of RC delay, which requires more power to drive those signals. It also produces heat, which increases electromigration and accelerates circuit aging.

This is one of the reasons why advanced packaging has begun gaining traction, particularly 2.5D and fan-outs, which also can take advantage of the fact that analog circuitry doesn’t benefit from node shrinks. Using logic libraries with lower levels of leakage will prevent frequent reloads and save power, Lowman said. Adding the ability to do some trigonomic functions with a digital signal processor also allows an application to verify its own environmental readings in a single cycle rather than several.

If the sensor has enough compute power run some calculations of its own, that also could eliminate the need to send data to the cloud so frequently. The time it takes for a processor to calculate for a few cycles is much less than the cost in power to wake up a device, keep it awake to send data and wait to receive a response.

The more processing power a remote device has, the more work it can do itself and the less it has to send to the cloud, Miller said.

“If you don’t have a low-cost gateway nearby, that may not work, but if you have enough processing on the edge device, a little computation can save you a ton of power,” Miller said.

Whether that energy harvesting will alter this equation isn’t clear at this point. Light, heat and motion can still generate ambient energy. Fraunhofer IIS’s BlueTEG module, for example, uses the temperature difference between a heat source and the environment to monitor condition and structural health.

Fig. 1: Fraunhofer’s BlueTEG module in action. Source: Fraunhofer IIS

And Georgia Tech has produced an array of 1,000 LEDs that can be powered by the force of a shoe striking the floor.

Fig. 2: Georgia Tech professor Zhong Lin Wang poses with an array of 1,000 LED lights that can be illuminated by power produced by the force of a shoe striking a triboelectric generator placed on the floor. Photo: Georgia Tech/Rob Felt

But when all of this harvesting technology will be necessary to either augment batteries, or potentially eliminate them, isn’t obvious at this point. In the end, it will come down to price—a function of mass adoption and more standardized approaches—as well as proven reliability and consistency. Right now, batteries and efficient chip architecture are the real competition, and there appears to be more room to improve the efficiency of designs and the compute power within a given energy budget before additional componentry is required.

Related Stories
Energy Harvesting Gains Steam
Shifting process nodes and new applications are driving demand, but utilizing this technology requires a different design mindset.