DARPA’s near-zero-power sensor technologies have extended battery lifetimes from four weeks to up to four years, but more work needs to be done...
Some battlefield sensors that used to run out of power in months (if not weeks) can now keep providing valuable intelligence for up to four years before their coin batteries need to be replaced. That’s one of the more dramatic results of DARPA’s Near Zero Power RF and Sensor Operations (N-ZERO) program, which concluded in May 2020.
DARPA began the initiative in 2015 with a goal of finding ways around the lifetime limitations of IoT battery power so that sensors could be deployed in the field to detect events like vibration, light, sound, or other battlefield signals without having to be frequently replaced.
Battery lifetimes can now be extended by using sensors that only wake up to record a triggering event, such as a sound, or to periodically (instead of continuously) monitor a battlefield or other concern, said DARPA.
Troops are now able to gather intelligence without having to venture into a potential combat zone themselves. They also have the reassurance that they would be moving into hazardous areas far less often to replace “battery-dead” sensors.
“The N-ZERO program established asleep-yet-continuously-alert sensing capabilities for untethered, unattended systems that are triggered by specific physical or radio frequency [RF] signatures,” said Benjamin Griffin, program manager in DARPA’s Microsystems Technology Office. “Sensor lifetimes that are on the order of years will enable cost-effective and safe deployment of sensing technologies in areas lacking fixed-energy infrastructure.”
Sensor technology improvements are ongoing, with sensor capabilities likely to expand. What Griffin acknowledged for now was that RF, acoustic, and infrared (IR) wake-up capabilities were the most successful sensing systems developed by the N-ZERO program. These near-zero-power sensors transmit radio signals to communicate (RF), measure sound levels (acoustic), and detect IR (thermal) radiation.
“What pleased us most was the power consumption of the best performers, which resulted in an estimated battery lifetime extension from four weeks to up to four years, based on a coin-cell battery,” Griffin explained.
Despite these successes, the battery lifetimes of sensors tested in the N-ZERO initiative are still limited by processing and communications of confirmed events and/or ultimately by the self-discharge of the battery itself.
One move in a positive direction was the development of an ultra-low-power Arm processor. The Arm M0N0 processor achieved a 10-nW sleep power and 20- to 60-µW/MHz active power level, depending on the application. Sensor power shutdowns allow users to store information in read-only memory (ROM) that can be accessed without the power penalty of recovering the data from off-chip non-volatile memory.
Arm chips also have the capability to run for decades on one set of batteries, as their power consumption is so low. In comparison, a conventional coin-cell battery in sleep mode might have a lifetime of multiple years, but it can’t operate much beyond that window.
“This is important because troops need to know about enemy movements and changes to the battlefield,” said Griffin. “However, since sensors constantly consume power, they often spend time processing what turns out to be useless data. That power consumption then saps the sensors’ battery lives.”
DARPA said that lower processing options that dramatically increase battery life (like Arm) can cut costs that get incurred when sensors must be replaced frequently because their batteries fail. Safety is also enhanced because military service personnel no longer have to physically travel to dangerous areas to replace in-field sensors that have failed and need to be replaced.
“An example implementation was a neural network implemented as a keyword spotter that processed audio files from the Google Speech Commands dataset at a power level that could run continuously for 200 days on an LR44 coin-cell battery,” said Griffin. “Working with DARPA, Arm has made 1,000 licenses available to the U.S. government and government contractors to allow for immediate access to processing in power-constrained environments.”
At the same time, solid-state sensors are beginning to appear in the defense marketplace. These sensors were not part of the original N-ZERO program, but they are being tested in the Department of Defense (DoD) applications.
“We have worked with the DoD and with DoD subcontractors in the area of low-power sensors,” said Jeff Sather, vice president of technology and customer solutions at Cymbet, a producer of very small solid-state batteries.
“They want to minimize sensor components and also be able to use alternative sources of battery recharging such as solar energy,” he said. “What we do is integrate very small solid-state batteries together with the sensors on a single chip. With solid-state batteries, you have no battery leakage and no risk of sparks or fires, but the cost of these integrated, solid-state battery, low-power sensors isn’t cheap. It is around $1 per sensor, when you can get a low-power coin-cell sensor that uses a traditional liquid lithium battery for around 25 cents.”
Price points for near-zero-power sensors integrated with solid-state battery technology are likely to come down, which could make solid-state sensors the next popular low-power sensor option, but for now, the DARPA N-ZERO program is primarily using coin-cell batteries.
DARPA’s N-ZERO program results are being expanded to a broader field of applications.
“It is anticipated that N-ZERO’s largest use case will be within the IoT, where the power consumption of sensor nodes is one of the technical challenges to pervasive expansion of the technology,” said Griffin.
“In the area of agricultural and structural health monitoring, ARPA-E has continued funding under the OPEN+ Sensors for Bioenergy and Agriculture Cohort to develop sensors for plant monitoring,” he added. “Other transitions of the N-ZERO technology have focused on untethered health monitoring of mechanical systems.”
Thus far, according to Griffin, the majority of successes within N-ZERO were enabled by sub-threshold circuit design.
“Successful sub-threshold approaches stored signature data in ROM or in volatile static random-access memory (SRAM) because these technologies have low-read energy requirements,” he said. “The use of ROM constrains applications because the memory is set during fabrication and cannot be reprogrammed. Since SRAM is volatile, it loses memory if there is a drop in the bias voltage to the IC.”
High on Griffin’s “wish list” is a low-read-energy, non-volatile memory that would give untethered systems the ability to be re-programmable while simultaneously making them robust to power loss.
This wish coalesces with Sather’s opinion that the next focus of low-power sensor technology is likely to be in the area of memory. “We know that the storage capacity on these devices will need to be increased and that there is a need for continuous improvement,” he said.