Energy Harvesting Makes Progress

The dream of a self-powered device has been around for a long time—thousands of years, in fact. The first windmills and waterwheels date back to ancient Greece and the beginning of recorded history. Self-winding timepieces date back to the late 1700s. The Foucault pendulum has been in motion in Paris, minus some brief hiatuses, since 1851. And it’s been 144 years since two French physicists discovered piezoelectricity, in which a charge remains inside certain materials when mechanical force is applied.

For the most part, these approaches to generating energy fall into the camp of extremely bulky but effective, as in the case of wind, solar or water power, or they are limited to what basically amounts to scientific experiments. But as the need grows for more functionality in tiny portable devices, particularly for wearable and implantable electronics—and as the gap between battery improvements and IC functionality continues to widen—energy harvesting is beginning to attract serious attention and investment.

“In practice, energy density from these sources is still very low, best-case in the microwatt range for portable apps, though you can get higher from industrial machine vibrations, for example” said Bernard Murphy, chief technology officer at Atrenta. “More typical is tens to hundreds of picowatts. So it’s a long way from powering a smart phone, which needs milliwatts, or any kind of wearable comparable to a smart phone.”

As with most complex designs these days, there is no magic bullet. A combination of dark silicon, complex switching schemes, fewer and better-defined functions in some cases, more power-efficient architectures and better software-hardware co-design are all still required. Lower leakage, new materials, and approaches such as Wide I/O-2 or through-silicon vias to reduce the amount of power needed to drive a signal, are also under consideration. But being able to incrementally add battery life whenever possible doesn’t hurt, either.

“If there’s spiky power demand (mostly off, occasionally on), charging a capacitor/super-capacitor can work,” said Murphy. “Energy from ambient wireless falls off as the square of distance from the source, so that’s hopeless unless you can guarantee you remain close to the source, which may be possible with industrial applications. Photovoltaics and thermoelectric are in the tens of microwatts per square centimeter range. And ambient wireless falls off with square of distance from source, so negligible value for mobile/wearable/implants.”

He noted there also is work under way to increase power density from piezo/flexo approaches through MEMS structures, whereby many bending cantilevers could potentially generate power in higher densities. One such generator has been able to produce nearly 0.8 milliwatts per square centimeter in a lab, which is getting close to phone levels.”

New demand in new places
Not everything will require even that level of power, though. But as more devices become mobile and connected, there definitely will be more batteries to charge. More electronics, even in the home, have at least a battery backup rather than just a plug.

“If you’re going to have 1,000 things in your house and you have to change a battery every year, that’s an average of three battery changes a day,” said Rob Aitken, an ARM fellow. “What are the alternatives? Energy scavenging, 10-year batteries, or a battery service that comes to your house once a year to change out all the batteries.”

Energy scavenging, also known by the more marketing-friendly term, “energy harvesting,” is the option that has garnered the most interest because it’s the least disruptive to lifestyle, and it can be used to supplement batteries that are hard or inconvenient to replace. And even within the current energy-harvesting limitations, it can be useful if the chip architecture is modified.

“Because the power is in the microwatt range, you have to change the way you do computing,” said Aitken. “One alternative is that you can run something all the time. It never sleeps, and it consumes 80 microwatts continually. That is somewhat of a change in the compute model, but it’s not inconceivable. You can have a microcontroller running at that kind of a level, and as long as you limit the communication so it doesn’t use up all of the energy communicating, then you can interact with other things. It’s within the existing computational models, but it’s a slightly different model than we use now.”

There are other twists on existing models finding their way into the market these days, as well, as real commercial products. Maxim Integrated, for example, just introduced a secure NFC/RFID tag authenticator with built-in energy harvesting aimed at the medical and industrial markets. What’s different about this approach is where the tags get their power.

“The reader sends a signal to the tag, the tag then recovers that energy from the wave coming to it, powers up, and uses that wave as a means of communication,” said Hamed Sanogo, executive business manager at Maxim Integrated. “We’ve reliably cleaned up the RF front end so that the wave goes into the part. The energy needed for charging this is in the microwatt range, while the reader devices are in the milliwatt range.”

Sanogo said that in the past the RFID/NFC tags would simply dump the excess energy, but with the focus on extreme low power, particularly in medical applications and sensors, every microwatt counts. In the future, that reader may be a smart phone rather than a specialized device, too, and the energy sent to one tag may be used to wake up other tags in a neural network.

Maxim Integrated’s energy harvesting architecture.

Design considerations
From an IC design standpoint, energy harvesting adds a couple of twists. One of them involves partitioning. What happens, for example, when communication may be to a nearby device rather than a base station located hundreds or thousands of yards or meters away?

“One of the big changes is in partitioning and placement,” said PV Srinivas, senior director of engineering for the Place & Route Division at Mentor Graphics. “You have to make sure that you can communicate with your neighbor. Timing constraints have to be taken into account, and so does communication density. What voltage is available where? Is it bufferable or non-bufferable? Placement becomes a very critical factor.”

So is cost. Being able to reduce voltage and generate energy through a variety of means is essential, but it also has to be done at a low cost because many of the applications that will benefit from energy harvesting are price-sensitive.

“The objectives for these designs are low cost and low power, and you cannot afford to compromise one for the sake of the other,” said Krishna Balachandran, product marketing director for low power at Cadence. “You want to enable energy harvesting, but you also have to keep the cost down when you do it.”

Other options
While we tend to think of energy harvesting in terms of kinetic or solar energy, there also is work under way in other areas. One involves biology. As early as 2008, a group of researchers from MIT published a paper in 2008 that looked at the sustained voltage difference between plants and their surrounding soil.

“Some interesting approaches in microbial fuel cells to generate through anaerobic fermentation process are projected to be able to produce up to 2KW on an industrial scale,” said Atrenta’s Murphy. “Implantable fuel cells may be more likely to work using enzyme-based oxidation of blood sugars and have been demonstrated at around 2mW/cm². But you do have to change the battery every two years, which is the downside.”

MIT also is working on microengines on chips, and Darpa several years ago was experimenting with chip-sized nuclear reactors. What ultimately becomes commercially viable is unknown, but research is beginning to ramp as design teams working on ever-greater functionality in smaller and smaller devices search for ways to sidestep the limitations of conventional battery technology.