See If Today’s Air Quality Will Conch You Out

Air quality is one of those problems that is rather invisible and hard to grasp until it gets bad enough to be undeniable. By then, it may be too late to do much about it. But if more people were interested in the problem enough to monitor the air around them, there would be more innovators bringing more ideas to the table. And more attention to a problem usually means more accountability and eventual action.

This solar-powered particulate analyzer made by [rabbitcreek] is a friendly way to take the problem out of the stratosphere of ‘someday’ and bring it down to the average person’s backyard. Its modular nature makes it fairly simple to build, and the conch shell enclosure gives it a natural look. That shell also cleverly hides the electronics, while at the same time allowing air and particulates to reach the sensor. If you don’t like the shell enclosure, we think the right type of bird feeder could protect the electronics while allowing airflow.

[rabbitcreek] attached a sizeable solar panel to the shell on a GoPro mount so it can be adjusted to face the sun. The panel charges a Li-Po battery that gets boosted to 5V. Every two hours, a low-power breakout circuit wakes up the Feather ESP32 and takes a reading from the particulate sensor. [rabbitcreek] can easily see the data on his phone thanks to the Blynk app he created.

Why limit this to your yard? Bare ESP32s are cheap enough that it’s feasible to build a whole network of air quality sensors.

16 thoughts on “See If Today’s Air Quality Will Conch You Out

  1. I think this is a great concept and a super simple way to get some particulate matter data. The only issue I have with this Instructable is the same issue I see with many others like it: part list pricing and source. For some reason they show item #8 on the part list the Honeywell HPMA115S0 and then they randomly show the part number for a different sensor, the PMS5003. These are similar air particulate matter sensors (and are both the much more accurate laser IR type being laser IR vs focus IR LED) and I think they output data in the same fashion based on my experience, but the Honeywell sensor is much more reliable and costs 35$ vs the PMS5003 that is around 18$. The main issue with the PMS5003 is its lifetime and data reliability because it has so many different Chinese manufacturers. I understand that prices change over time and that sources can change as well, but it seems like the prices of some of these items aren’t quite right. Again, I don’t want to take away from this great idea and its simplicity, but for the sake of people out there trying to build these things its nice to see more accurate price information or at least explain why the prices of said items are so low for the people who aren’t electronics gurus to know why something costs basically 0$. Many teachers and students are trying to build these for their classrooms after all.

  2. Accuracy on this project is going to be a mess. PM sensors use a resistive heater and a fan, and have a settling period numbered in hours. To deal with battery life, he’s turning the whole system off between readings. That will make this possible to run via solar, but your readings are going to be way off because the sensor is no good until it has been powered on continuously for several hours.

    Currently, to the best of my knowledge there is no good solution for using harvested energy to power PM sensors unless you can generate pretty substantial amounts of power and store it all somewhere that can deal with ambient weather conditions (problematic for lithium chemistries in colder climates).

      1. I have a fleet of the PMS5003 sensors in the field collecting data for a couple years now. To test them, I placed a series of them near a calibrated EPA sensor (see here: https://imgur.com/a/joQvj) and then pulled data from the sensors and the FMR PM sensors. The chinese datasheet (which I did read, thanks for asking) notes that the settling interval from deep sleep is on the order of 30s to allow for fan spin up, but in my own experience they need several hours before they come into alignment w/ the FMRs.

        I have physically disassembled one of these and there appears to be a resistor being used for internal temperature stability, and conversations with PhDs in the field suggest common practise for laser scatter PM sensing is the same. The datasheet does not note this, so I may be completely incorrect on the working mechanism. I am however confident in the time interval and have years of data to back it up.

        I currently have several sensors offline with some new firmware being rolled out, but current measurements from a portion of the test fleet can be viewed at https://graqm.org

        1. That’s actually really cool that they’re letting you site your meter next to a bunch of calibrated meters.

          Do you know if the inaccuracy from off/on cycles is a constant offset, a multiplicative term, or just totally random? Because if it’s something that could be calibrated out, it would be super helpful to know, and you might be in a position to find out.

          1. We spent a fair amount of time with hopes of being able to calibrate these things beyond the default values presented by the device. Settling activity presented itself as noise, which would eventually drift closer to the value recorded by the continuous read PM sensors.

            The EPA sensors come in a couple of flavors. The devices of interest are continuous sampling, which we were able to pull data from every minute (which might be an artifact of their data collection). These are FEMs as opposed to FRMs which use a filter media that is pulled daily and sent to a lab somewhere in FL. The FEMs will generally dehumidify the incoming sample as relative humidity can have a serious impact on the readings.

            This is a little difficult to accomplish with an inexpensive sensor, and work has been done on the PLANTOWER devices used in the OP and in my own project showing the impact of sample humidity on accuracy, and it isn’t good: https://www.researchgate.net/publication/327251921_The_influence_of_humidity_on_the_performance_of_a_low-cost_air_particle_mass_sensor_and_the_effect_of_atmospheric_fog

            So, the primary issue here is understood to be ambient relative humidity impacting the readings, where higher humidity == mo’ problems.

            The hope here was to somehow use the magic of statistics to deploy inexpensive PM sensors along with ambient temp and humidity readings to develop a calibration curve which can be applied to the raw sensor data.

            However, I am the sole designer behind this project and while I wear many hats, data science ain’t one of them. The assumption is that this approach can work, but it’s going to require someone smarter than me. For the time being I have applied a linear offset to the O3 and PM2.5 readings reported, which is device-specific and was pulled from a comparison between each device and the FEM data recorded, using nothing more than an average over time after the device had settled for a day. Humidity is being reported and recorded, but is not factored into the readings I publish quite yet.

            In true hacker fashion, I’m choosing to ignore the problem and work on something I understand better, which currently is dealing with crappy GPS units from China occasionally telling the world that I have an air quality sensor somewhere in the north Atlantic.

        2. Laser physicist here..

          Depends.. laser scattering is pretty much instantaneous, and while the power output of any laser diode is temperature dependent, you ought to be able to characterize the power variation versus temperature as a correction factor for the readings. You’re more likely to have problems from inadequate averaging or a poorly implemented detector circuit.

          The sensor in question is in fact not from China. The author lists the manufacturer as Honewell (HONEYWELL HPMA115S0-TIR PM2.5), and while the device may be made in China, I’d like to think they already thought of this. Maybe not.

          Datasheet here: https://sensing.honeywell.com/honeywell-sensing-hpm-series-particle-sensors-datasheet-32322550-e-en.pdf

          The unit is also listed as being designed for harsh environments, with temp ranges spec’d at -20 to 70C (-4 to 158F). All this, with the max operating current specified at less than 80 mA (600 mA inrush, which is probably the start of the fan and maybe a bit of the laser).

          Unless you’re creating energy from nothing, 80mA isn’t gonna cut it to provide thermal stability over that range. More likely, the internal monitoring diode in the laser diode assembly is used either as feedback for stabilization, or as a signal for offset of the measurement to ensure accuracy.

          1. The HPMA1150SO and the Plantower PMS5003 are different sensors. The remark I made about them being chinese was about how there are many knock off versions of this sensor being made by companies that aren’t Plantower.

  3. Using PM sensors to qualify Air Quality is a bit of a misnomer.

    Most inexpensive PM sensors are in the range PM10 to PM 2.5

    Particles this large are certainly hazardous to your lungs however the things that can seriously effect your health are much smaller than this.

    Most things PM2.5 or larger can’t get any further than your lungs, while this in itself is not good for your lungs at least they are non-invasive and do not progress into your blood.

    Small things like VOC’s, Bacteria, Micro-organisms, Gasses and mycotoxins for example can be invasive and cause serious debilitating health issues, if not death.

    I think we should let go of high precision, high accuracy and hence costly sensors and instead attempt to characterize the air.

    Air Quality is not a random selection of all permutations. Specific environmental elements that exist in many places will yield the same air quality issues.

    A variety of poor precision and poor accuracy sensors is likely to be far more effective at accurately characterizing air quality then a small number of expensive sensors especially if they are used collectively to gather large data-sets.

    An example would me mycotoxins, these are invasive and once they enter your blood system they can cause havoc. 25% of people (through their personal genetics) do not have the biological ability to remove these toxins and they suffer debilitating illness. Many mycotoxins are also neurotoxic.

    Mycotoxins are in the range PM0.1 to PM0.4 so the only current way to test is with Polymerase Chain Reaction (PCR) DNA amplification with is an expensive scientific process (ERMI).

    However Mycotoxins (as the name suggests) come from one member of the Fungi group, that is, Mold (Mould). These mycotoxin producing molds are present in water damaged buildings. Note: only some specific molds produce mycotoxins.

    These molds are in the range PM2.5 to PM5.

    A range of sensors including PM2.5, humidity, condensation point, temperature, color temperature would likely be able to characterize an environment that is high in dangerous mycotoxins without needing a highly expensive process to measure their presence and indeed not measuring them directly at all. And at very little comparative expense.

    There is one German organization that is compiling a so called Air Quality national database but unfortunately they rely extensively (almost exclusively) in PM sensors.

    It would be great to see someone take this a step further with an inexpensive multi-sensor open hardware product and an open source sharable IoT network. This might be something that IoT is actually good for.

    1. I haven’t seen one of those multisensors you are talking about, could you perhaps share a link for one? I’d be really interested in purchasing one for a future project implementing one!

      Also thank you to everyone discussing this and informing me about this subject more!

Leave a Reply to MedixCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.