BBC London ran a recent story on air pollution exposure during commuting:
It now seems a long while since our last blog post, but air pollution is barely out of the news with the latest BBC story about air pollution exposure during commuting. This topic is fairly close to our hearts with the pollution guardian activity inspired by the fact that air pollution exposure is highest within motor vehicles.
Earlier on this year, we put together our first prototype pollution sensor containing a particulate sensor for PM2.5 measurement and a Nitrogen Dioxide (NO2 ) gas sensor. PM2.5 refers to the measurement of very small particles, less than 0.0025mm in diameter which can go deep into the lungs and containing a variety of payloads e.g. road, brake or tyre dust or chemical compounds from combustion. Nitrogen Dioxide gas can exacerbate respiratory conditions and at street level is mainly produced by diesel vehicles.
How does out first prototype look?
Why did we make this first prototype?
to create a self-contained sensor unit with its own internal power supply to enable easier bench and field (car) testing
to shrink the volume of air trapped within the unit in order to keep the unit responsive to the current air conditions
to enable experimentation with parameters to find the best balance of sensitivity versus stability
We will write a bit more about the “infrastructure” of the prototype and surrounding app and cloud solution in later blogs, but the main question we had when we put the hardware unit together was does it work?
Well, the good news was that the prototype actually worked as expected across a range of functional areas with only one bug with battery charging which was solved by a simple hardware modification to the units.
How well did the units work compared to our initial cardboard prototypes? Pretty good in practice:
having “always on” internal power meant that the sensors could be ready to start work within a couple of seconds; previously the sensors could require several hours from being powered up to being ready to measure
fast responsiveness to Nitrogen Dioxide gas e.g. able to pick out smelly vehicles passing by
And where does that spike of air pollution correspond to? A smelly van on the southbound slip road of the M25/A3 junction…
As mentioned in the previous blog, we put in a lot of preparation work over summer 2018 in order to be able to quickly start work on the Pollution Guardian project and de-risk certain areas. One of the first items to be actioned was the creation of a custom circuit board to explore the performance of our selected, affordable Nitrogen Dioxide (NO2) gas sensors and the type of electronics needed to setup, amplify and filter the output of these sensors. We called it our “bench test” circuit board as we thought it would only get used indoors on the bench..
With the bench test board, we designed a couple of options for the electronics solution:
to provide a backup option, in case the one circuit failed to perform
provide us alternate options based on relative performance
to enable optimisation of the bill of materials cost
Designing & making the circuits went relatively smoothly, but it is not until you receive circuits that you can really see what is going on. Some mistakes we discovered quickly, whilst others took a period to debug. Mostly our mistakes were due to the misinterpretation of and assumptions on the component specifications; our first solutions to these issues actually went on to compound our problems.
After a period of investigation and debugging, we were finally able to identify and solve the underlying issue. As part of this work, we lashed these prototype circuits together with an off-shelf development board (TI cc2640R2) all contained in a cereal packet wrap for protection. We called our first units Dougal & Zebedee and took them out of the lab and into the car:
What we found when we started testing out the units was that whilst our units appeared quite sensitive to temperature change, they were actually capable of detecting low levels of NO2 inside the car.
The picture below shows a trace of NO2 measurements whilst traversing the one way gyratory system in the centre of Guildford – note that a decreasing level on the graph trace implies a higher NO2 concentration:
So, at the conclusion of our cardboard proto work, we knew we had the right potential sensitivity toward NO2 which we could work with on the next prototype to calibrate and try to get consistent between test units.
Feasibility projects like the Pollution Guardian imply a certain level of risk and it was critical for our business to secure some grant support from Innovate UK to help us mitigate those risks.
Looking back, it seems a long time since we made our application for funding, but considering the risks ahead, we prepared ourselves in order to get started quickly in case of a positive outcome of our bid:
Working on the system architecture, driving decisions on the hardware and software platforms to use within development
Further research on the key components
Talking to collaborators and contractors to re-check availability
Re-planning the market research
Now, many academic papers point to the difficulty in using affordable sensors e.g. around variability, stability and accuracy, so one of our biggest challenges in the project is to put together an affordable solution based on these sensors. Our approach was to tackle this issue head on, building a very early prototype and using a minimum “data gathering platform” around it to understand the pitfalls & performance issues.
This early work cut across several disciplines:
Electronics, designing a custom circuit to best interface with affordable gas sensors
Firmware, building on a development platform to gather and share sensor data over wireless
Mobile app, customised to gather the local sensor data & upload it to a data store
Mobile backend, a real time database to capture the data streams and tools to help explore the data
Mechanics, how to wrap this early prototype for real world use
Early system testing & validation approach
We will be adding a few blogs to the site to cover progress on the above items; suffice to say the first mechanical housing for the unit followed the “breakfast cereal box” approach. After wrapping up the first sensor unit, its looks gave us the name for our prototype, Dougal, after the dog character from the programme, “The Magic Roundabout”.