News

Researchers Sniff Out Common Pitfalls of “Electronic Noses”

February 02, 2021 by Adrian Gibbons

Sensor R&D has targeted all of the five human senses—including smell. Here are a few ways "electronic noses" have advanced even further in recent months.

Beginning in the early 20th century, canaries were an important sentry for detecting carbon monoxide and other harmful gases in mines. This tradition underscores the necessity to detect hazardous fumes in working environments. 

Nearly thirty-five years later, olfactory technology is revealing not just the presence of a noxious odor but also the concentration and differentiation of similar gaseous compounds. Last year, for instance, Intel offered a peek at an enhanced machine intelligence chip that learns smells.

 

Intel's Loihi neuromorphic chip

Intel's Loihi neuromorphic chip contains artificial neural networks that mimic the way the human brain reacts to smell. Image used courtesy of Walden Kirsch and Intel 
 

Other researchers and engineers have recently continued this pursuit, developing "electronic noses" to create safer working environments and better quality assurance in food production. 

 

How Do Electronic Noses “Smell?”

Before delving into these new research advances, it may be helpful to first review the basic technology that underlies electronic noses. Traditional methods for identifying and detecting odorous gases, chemical-specific analytics, and olfactometry, are typically expensive and do not operate in real-time. This limits opportunities for mass IoT adoption.

According to Odotech, electronic noses consist of three elements:

  1. Array of gas sensors
  2. Data pre-processor
  3. Data interpolation engine

These elements are electronics-based in many cases and mimic the way that the human olfactory sense works. Sensors on a chip will modify specific electrical characteristics, changing in the presence of the select gases.

 

Human olfactory system vs. electronic nose

The technology is modeled similarly to the human brain. Image used courtesy of Odotech and ResearchGate
 

In 1953, a metal-oxide semiconductor was observed changing its resistance profile in the presence of oxygen, kicking off early sensor research. Now, the subsystem in e-noses responsible for detection is made from various transducers including metal-oxide semiconductors (MOS) and MOSFETs.

 

Start-Up Offers Multi-Channel Detection in a 1 mm2 Chip

SmartNanotubes Technologies is a start-up looking to revolutionize IoT applicability of the electronic nose. The detector chip, the “Smell iX16” is a 16-channel array, sampling signals at a rate of 1.8 seconds, according to the company. 

The prototype PCB, the “Smell Board iX,” runs four of these detection devices, allowing up to 64 channels worth of odor detection. The detection process operates by characterizing “patterns” that relate to different scents. 

 

 Device observing the presence of an orange by smell

SmartNanotubes’ demonstration at the all-virtual CES 2021 shows its device observing the presence of an orange by smell. Image used courtesy of SmartNanotubes Technologies and DigitalTrends
 

SmartNanotubes Technologies suggests that many gases can be detected and characterized with their 64-channel system. Currently, they claim to have successfully detected several compounds including ammonia, carbon dioxide, isopropanol, and bananas.  

 

Skoltech Develops Electronic Nose Using Additive Manufacturing

Using additive manufacturing techniques, SkolTech has likewise developed a PCB e-nose with the aim of reducing the cost of sensor technology and pushing it to commercialization. 

The precision of additive manufacturing has reached a level where “the resolution of the printing is close to the distance between electrodes on the chip, which is optimized for more convenient measurements,” according to senior researcher Fedor Fedorov. 

He goes on to elucidate the benefits of the printing process: “We managed to use several different oxides, which enables more orthogonal signal from the chip—resulting in improved selectivity.”

 

This e-nose matrix board contains eight sensors

SkolTech’s proof of concept PCB with eight sensors. Image used courtesy of SkolTech
 

SkolTech indicates that the sensor is able to differentiate between chemically similar compounds including methanol, ethanol, isopropanol, and n-butanol at low concentrations—which could be a key safety feature in detecting poisonous alcohol mixtures in food settings.

 

The multiple sensor arrays

The multiple sensor arrays were developed by 3D-printing nanocrystalline films of various metal oxides and bonding them onto a chip. Image used courtesy of ACSPublications
 

SkolTech researchers mention one drawback of the device: currently it only operates at temperatures between 200–400°C. The researchers hope that subsequent materials research—potentially the quasi-2D material called MXenes—might provide fabrication materials that would operate at room temperature. 

 

Biotechnology Hybridization Allows For Detection of VOCs

Researchers led by Professor Shoji Takeuchi from the Biohybrid Systems Laboratory University of Tokyo have developed a cell-free (as opposed to cell-based) hybrid volatile organic compound (VOC) sensor

The researchers attempted to measure the presence of Octenal (a volatile organic compound present in human sweat) by combining receptor proteins from mosquitoes. To achieve this, they machined 16-channel microtubes for moving the gases through the system.

 

A detailed view of one channel of the system with droplets added into the two wells to form a lipid bilayer

A detailed view of one channel of the system with droplets added into the two wells to form a lipid bilayer. Image used courtesy of UTokyo
 

The presence of Octenol was measured electrically by sampling current at a rate of 5 kHz through a 1-kHz Bessel low-pass filter with a voltage reference of +60 mV. The overall goal of the project was to utilize the olfactory techniques of living organisms in the detection of VOCs with precision and selectivity far greater than current VOC sensors. 

 

Teaching Chips to Smell

While chips don't exactly "smell" in the way humans do, these up-and-coming technologies do ameliorate common machine learning, manufacturing, and design-level challenges associated with electronic noses. In effect, e-nose sensors may one day be a common feature in IoT devices to keep homes and workplaces safe. 

 


 

Do you work in a field where the sense of smell could be automated, thereby improving safety or quality? Let us know in the comments below.