Researchers at Penn State have come up with a paradoxical way of improve weak signals from sensors: Adding in a small amount of white noise.
The team's work centers around a next-generation, ultra-low-power photodetector, constructed from a two-dimensional layer of molybdenum disulphide, designed to detect light — but, the paper suggests, the same concept can be extended to other sensor types, using a fraction of the energy and space required of the sensors currently deployed for the Internet of Things (IoT).
Traditionally, background noise is something to filter out of sensor readings — but the team's work found turned to nature for inspiration in how background noise can actually be used to improve sensor readings: Stochastic resonance. "Stochastic resonance," explains co-first author Akhil Dodda of the work, "is a phenomenon where a weak signal which is below the detection threshold of a sensor can be detected in the presence of a finite and appropriate amount of noise."
"This phenomenon is something that is frequently seen in nature," adds Saptarshi Das, assistant professor of engineering science and mechanics. "For example, a paddlefish that lives in muddy waters cannot actually find its food, which is a phytoplankton called Daphnia, by sight. The paddlefish has electroreceptors that can pick up very weak electric signal from the Daphnia at up to 50 meters. If you add a little bit of noise, it can find the Daphnia at 75 meters or even 100 meters. This ability adds to the evolutionary success of this animal."
In the paper, the team was able to demonstrate that by adopting stochastic resonance and adding Gaussian white noise to the signal. While doing so costs energy, the team measured it at mere nano-Joules — and found that the photodetector became able to detect weak signals from a distant LED which were otherwise below its detection threshold.
The team's work has been published under open access terms in the journal Nature Communications.