Power/Performance Bits: Feb. 14

Electronics for Venus

A team of scientists at NASA’s Glenn Research Center in Cleveland demonstrated the first prolonged operation of electronics in the harsh conditions found on Venus.

Current Venus landers can only operate on the planet’s surface for a few hours due to the extreme atmospheric conditions. The surface temperature on Venus is nearly 860 degrees Fahrenheit, and the planet has a high-pressure carbon dioxide atmosphere. Because commercial electronics don’t work in this environment, the electronics on past Venus landers have been protected by thermal and pressure-resistant vessels. These vessels only last a few hours, and they add substantial mass and expense to a mission.

To overcome these challenges, the Glenn team developed and implemented extremely durable silicon carbide junction field effect transistor (JFET) ring oscillator integrated circuits. They then electrically tested two of these integrated circuits (up to 24 transistors, with two levels of metal interconnect) in the Glenn Extreme Environments Rig, which can precisely simulate the conditions expected on Venus’ surface. The circuits withstood the Venus surface temperature and atmospheric conditions for 521 hours – operating more than 100 times longer than previously demonstrated Venus mission electronics.

Integrated circuit before (above) and after (below) testing in Venus atmospheric conditions. (Source: NASA)

“We demonstrated vastly longer electrical operation with chips directly exposed — no cooling and no protective chip packaging — to a high-fidelity physical and chemical reproduction of Venus’ surface atmosphere,” said Phil Neudeck, lead electronics engineer for the work. “And both integrated circuits still worked after the end of the test.”

“This work not only enables the potential for new science in extended Venus surface and other planetary exploration, but it also has potentially significant impact for a range of Earth relevant applications, such as in aircraft engines to enable new capabilities, improve operations, and reduce emissions,” said Gary Hunter, principle investigator for Venus surface electronics development.

Low-power voice recognition

MIT researchers built a low-power chip specialized for automatic speech recognition. Whereas a cellphone running speech-recognition software might require about 1 watt of power, the new chip requires between 0.2 and 10 milliwatts, depending on the number of words it has to recognize.

In a real-world application, that probably translates to a power savings of 90% to 99%, which could make voice control practical for relatively simple electronic devices. That includes power-constrained devices that have to harvest energy from their environments or go months between battery charges.

The team focused the chip’s circuitry on implementing speech-recognition neural networks as efficiently as possible and optimizing the voice activity detection to prevent false wakes of the more complex speech-recognition circuit.

While the team tested three different activity detection circuits, the most complex of the three circuits led to the greatest power savings for the system as a whole. Even though it consumed almost three times as much power as the simplest circuit, it generated far fewer false positives; the simpler circuits often chewed through their energy savings by spuriously activating the rest of the chip.

Additionally, they minimized the amount of data the neural network chip had to retrieve from off-chip memory.

A node in the middle of a neural network might receive data from a dozen other nodes and transmit data to another dozen. Each of those two dozen connections has an associated weight which indicates how prominently data sent across it should factor into the receiving node’s computations. The first step in minimizing the new chip’s memory bandwidth was to compress the weights associated with each node. The data are decompressed only after they’re brought on-chip.

The chip also exploits the fact that, with speech recognition, wave upon wave of data must pass through the network. The incoming audio signal is split up into 10-millisecond increments, each of which must be evaluated separately. The MIT researchers’ chip brings in a single node of the neural network at a time, but it passes the data from 32 consecutive 10-millisecond increments through it.

If a node has a dozen outputs, then the 32 passes result in 384 output values, which the chip stores locally. Each of those must be coupled with 11 other values when fed to the next layer of nodes, and so on. So the chip ends up requiring a sizable onboard memory circuit for its intermediate computations. But it fetches only one compressed node from off-chip memory at a time, keeping its power requirements low.

The chip was prototyped through TSMC’s University Shuttle Program.

Cheaper, safer flow batteries

Researchers from Harvard University developed a new redox flow battery that stores energy in organic molecules dissolved in neutral pH water, allowing for a non-toxic, non-corrosive battery with a long lifetime. It could also significantly decrease the costs of production.

Flow batteries store energy in liquid solutions in external tanks — the bigger the tanks, the more energy they store. Flow batteries are a promising storage solution for renewable, intermittent energy like wind and solar but today’s flow batteries often suffer degraded energy storage capacity after many charge-discharge cycles, requiring periodic maintenance of the electrolyte to restore the capacity.

By modifying the structures of molecules used in the positive and negative electrolyte solutions, and making them water soluble, the team was able to engineer a battery that loses only one percent of its capacity per 1000 cycles.

“Lithium ion batteries don’t even survive 1000 complete charge/discharge cycles,” said Michael Aziz, professor of materials and energy technologies at Harvard.

The key to designing the battery was to first figure out why previous molecules were degrading so quickly in neutral solutions, said Eugene Beh, a postdoctoral fellow at Harvard. By first identifying how the molecule viologen in the negative electrolyte was decomposing, Beh was able to modify its molecular structure to make it more resilient. Using the same process, the team was able to utilize ferrocene, which while good at storing charge is normally insoluble.

The neutral pH should be helpful in lowering the cost of the ion-selective membrane that separates the two sides of the battery. Most flow batteries today use expensive polymers that can withstand the aggressive chemistry inside the battery, which can account for up to one third of the total cost of the device. With essentially salt water on both sides of the membrane, expensive polymers can be replaced by cheap hydrocarbons.

“Because we were able to dissolve the electrolytes in neutral water, this is a long-lasting battery that you could put in your basement,” said Roy Gordon, professor of chemistry and materials science at Harvard. “If it spilled on the floor, it wouldn’t eat the concrete and since the medium is noncorrosive, you can use cheaper materials to build the components of the batteries, like the tanks and pumps.”