New Sensor Hardware from the N4N Air Quality Nerds

Reblogged from Nerds For Nature.

When the EPA wants to know what’s going on with air pollution, they install a sophisticated sampling and analysis station that can cost upwards of a million bucks. Or they roll out one of their their high-tech air-sleuthing vans, crammed with more specialized mobile electronics than your typical Bond film.

Which is all very impressive, but what if I have asthma or a heart condition and want to know about the pollution on my street? What if you live next to a factory and want to collect data for further investigation? What if we are students or scientists trying to comprehend how pollution propagates in our communities?


Shaun’s Bluetooth-enabled mobile gas sensor board. — photo by Shaun Houlihan



Unfortunately, air quality sensors that are both relatively inexpensive AND accurate seem to be rare or non-existent. And not without reason: it’s a tricky task. Of course, new technology + inspiration = new opportunities. You may have heard about some of our electronics design and prototyping of less expensive, DIY-type air quality monitors to help address some of these needs.

To kick things off last year, Nature Nerd Shaun Houlihan developed an impressive PIC-based, Bluetooth-enabled, lithium-ion battery-powered toxic gas sensor board that can communicate with nearby smartphones and laptops. Nature Nerd Peter Sand is also testing a new air quality sensor board that uses a much cheaper (though much less accurate) metal oxide-based air quality sensor for use in his ManyLabs educational environment. You’ll be hearing more about this project soon.


L to R —> DD Scientific CO sensor, Alphasense reference CO sensor/board, and the new e-chem sensor board on the bench (with scope probes attached) — photo by Ken McGary

What’s New?

And we now have a third brand-new air quality sensor prototype board to tinker with! Nature Nerd Ken McGary just got our new e-chem toxic gas sensor AFE (Analog Front End) board debugged on the bench and roughly calibrated with some Arduino test code, and we have some fun things to show you about our progress so far.

The board was designed by Shaun, in collaboration with Ken and Peter — who successfully assembled the surface-mount parts onto the boards using his retro-tech “hot plate and a cheap skillet” technique. This new board is intended as a test and development platform for more refined and targeted sensor designs in the future, and as an educational tool for those trying to learn more about these sensors.

What’s An E-Chem Sensor?

This new breakout board provides the electrical interface between a microcontroller (like an Arduino board) and an electrochemical (e-chem) gas sensor cell. These sensors produce a tiny current in a diffusion medium by oxidizing or reducing the target gas. We are starting with carbon monoxide cells, but sensors are available for all sorts of species like carbon dioxide, nitrogen dioxide, ozone, and so on.


Internal structure of an electrochemical gas sensor cell (courtesy SGX Sensortech)



These are the same devices used in countless safety monitors for mines, airports, industrial environments, hospitals, homes, and anywhere else there is a potential hazard from elevated toxic gas levels. However, their use in lower-concentration atmospheric sensing applications is more recent, so some subtle effects on their output signals are not yet well understood.

They are also not “gas analyzers”, as they can be cross-sensitive to other gas species, and have other quirks that preclude absolute measurement certainty. But properly used they are impressive sensors that can provide reasonable gas concentration accuracy (within a few percent), sensitivity (down to parts-per-billion levels with some sensors and circuits) and reliability (with a typical service life of several years).


TI’s LMP91000 Analog Front End chip, the heart of the new sensor board, contains most of the analog circuitry for e-chem sensing — from TI datasheet.


What’s A Breakout Board?

After Shaun showed us his first working e-chem sensor board, we realized that there are other development platforms we’d like to experiment with the same gas sensors on as well — Arduino, RaspberryPi, and others. This wouldn’t be practical with his more specialized design, but Shaun’s gadget included one crucial teeny-tiny surface-mount interface chip that we still needed, the LMP91000 AFE from Texas Instruments.


Top and bottom copper layers of the e-chem sensor prototyping board — CAD image by Shaun Houlihan

So we decided to develop this new “breakout board” that includes the tiny AFE chip and some other tiny supporting devices (like a TI ADS1115 16-bit Analog-to-Digital converter), along with a clever socket layout that will accommodate several different sensor pin placements. After we worked out the schematic and parts list, Shaun laid out a nice circuit board form factor that can either plug into a standard prototyping breadboard, or can be used as a “Grove-style” sensor at the end of a 4-wire power/ground/I2C serial cable.


From analog (blue sensor signal out of the AFE chip at the top) to digital (serial clock and data signals on the I2C bus at the bottom), thanks to the ADS1115 A/D converter chip — digital scope capture by Ken McGary

Why Measure Carbon Monoxide?

Simply, to understand and potentially help diminish the complex stew of toxins and particulates that unfortunately find their way into our atmosphere and frequently our lungs. These pollutants come largely from vehicle exhaust pipes but also from industrial processes and power generation plants and even from the effluvia of our own consumer products being manufactured in other parts of the world.

Carbon monoxide is:

  • The most easily-found component of vehicle exhaust and industrial pollution, as typical atmospheric concentrations are in the ppm range rather than ppb. (EPA CO hazardous limits: 8 hours – 9 ppm, 1 hour – 35 ppm).
  • The least corrosive pollutant species and easiest/cheapest/safest to calibrate with using pre-mixed low-concentration gases (technically speaking, CO measurement is the “low-hanging fruit” of AQ sensing).
  • Colorless and odorless, yet deadly in moderately high concentrations, and potentially debilitating with chronic low-level exposure (see Zero to Million ppm CO concentration effects on the human body in this outdated but comprehensive chart).
  • Often a good proxy for other pollutants coming from the same sources (for example, NO2, SO2, fine particulates, and lots of other gunk spewed from diesel and gasoline engines).

How Are We Calibrating?

We are using a high-quality CO-B4 e-chem sensor from Alphasense [$115] for our calibration reference. It uses a recently-developed fourth “auxiliary” terminal to offset temperature and other sensor cell error signals. It is several times more sensitive than typical 3-terminal cells. This model is also much easier to calibrate in the system, as each unit is individually characterized at the factory and bar coded with the results — providing very good accuracy right out of the box.

We also purchased a matching Alphasense Individual Sensor Board (ISB)[$150] that provides amplified and filtered output signals, which are then directly measured by the same 16-bit A/D converter that is watching the AFE output from the sensor under test. Full-scale output for this ISB is 13ppm of carbon monoxide — impressive!


Dualing sensors in the cal chamber – Alphasense reference sensor using ISB Board (blue) and DD Scientific test sensor using new breakout board and preliminary calibration software — graph by Ken McGary
The other sensor you see in the graph is a less sensitive but much cheaper unit by DD Scientific [$38].There are several vendors of these types of devices, and a major goal is to characterize as many of them as we can get our hands on so that we can make the right cost/performance decisions for particular sensor applications. We are inspired by the efforts of Tim Dye and his associates at Sonoma Tech and beyond as well as the AirCasting researchers, and hope to soon be extending our understanding of these sensors in similar directions.


The Alphasense CO-B4 Individual Sensor Board (ISB), with an concentration accuracy of better than +/-1% and noise floor of around 10 parts per billion, is our carbon monoxide standard reference sensor, and costs about $300 including fees and postage from the UK. — photo by Ken McGary 

We’re also continuing our development of simple yet “pretty good” calibration rigs for e-chem and metal oxide sensors. We are starting with carbon monoxide and then we’ll tackle trickier species like nitrogen dioxide. We are using premixed cal gases, homemade test chambers, custom Arduino-based valve-metering equipment, and eBay-acquired flowmeters and gas regulators (more details on our new cal methods soon).

This straightforward and relatively inexpensive DIY approach will help us do our initial cal studies with some confidence without breaking the bank on fancy and expensive analytic instruments.This apparatus might also be used for “Cal Parties” and other community-based efforts to better calibrate and evaluate some of the cheaper sensor units that are out there, and could also be replicated for other groups to use in their own sensor projects.

So How Good Are These Sensors, Really?

To give you an idea of what these sensitive environmental sniffers can do, here’s some non-crucial yet sort of interesting preliminary data just from these sensors sitting on the bench. Sensor temperature (a few degrees higher than room ambient) and CO concentration was recorded every minute, starting around midnight.

The temperature data was mostly quiet until the timer on the house thermostat kicked in the burners for the central heat early on a chilly morning. Then the temp data starts to cycle as the furnace maintains temperature in the house. After a several-hour pause in the thermostat programming, afternoon heat kicks in again.

The CO data is pretty boring except for a couple of things: a) the large spike in the middle of the day corresponds to several doors opening up in the house, including the big garage door, so this shows the transient response of the sensor. The second item to notice is the modulation of the CO signal by the house heater in those tiny stair steps on the decay slope. These suggest a usable resolution on the Alphasense sensor of well under 100ppb (The spec sheet claims resolution of <10ppb!).


Sensor temperature signal (blue at top) shows mainly HVAC system cycling according to programmable thermostat. Alphasense CO sensor signal (bottom in red) shows large spike from outside (“fresh”) air incursion, then slow decay, with some modulation of CO levels by HVAC system — graph by Ken McGary

How Might We Use These Sensors?

We’re not going to be replacing the EPA vans anytime soon. But we think there are opportunities to develop moderately-priced yet “pretty good” DIY-oriented, well-tested and documented, and open-sourced:

  • Air monitoring stations for schools, towns, and community groups
  • Personal air quality monitors, dataloggers, and mappers for home, school, and work, various field investigations
  • Community-based calibration stations
  • Re-programmable smart sensor/datalogger boards and modules that can be economically adapted to new and novel uses.


The AirCasting project is an inspiring first step towards DIY air quality monitoring — image from 

One inspiration is the AirCasting project. It was developed as a platform for developing future mobile sensor projects and provides some interesting innovations and very nice (and recently upgraded) smartphone-based software.

Another interesting example is the EveryAware SensorBox project from Belgium, Italy, and the UK. The SensorBox design includes several electrochemical and metal oxide sensors along with an Arduino-based microcontroller core, forming a portable multisensor array or “e-nose” that reacts to traffic-related pollution. There are even efforts to use these to optimize traffic light patterns in real time, thus actually reducing pollution levels rather than just monitoring them.

So What’s Next?

Besides our sensor characterization and calibration efforts, we are also looking at options for commercializing some of these circuits so that other nature nerds can more easily tinker with these useful devices. And we’re looking for funding opportunities and research partnerships, as well, to help us design specialized apparatus and to further refine and document our DIY calibration methods. And we are eager to speak with any community, citizen action, or research group that can help us understand the real-world requirements for such devices to achieve actionable results.

We have the prototyping hardware, and a growing team with the skills and commitment to take the next steps, so let’s see what we can accomplish next! Maybe we can even come up with the AQ-monitoring equivalent to bioblitzing…

Join the N4N AQ Google Group or contact one of the team members to get involved.

Ken McGary for the N4N AQ Team

[Special thanks to Tony Trocian of SGX Sensortech for some of the information used in this blog post.]

Comments are closed.