air quality sensor (Plantower PMS5003) testing

[JBeale]

I don't know if this is really on-topic, as the PMS5003 just a small, cheap premade air quality sensor with serial I/O. Not much electronics to do yourself.  Anyway I was curious and tried one out, using an Arduino-type board to log output. It showed the air in my small office/workshop room is pretty clean, and doesn't change much with time for a few days. Then I happened to have a glue gun plugged in for about 10 minutes, on the other side of the room.  There was a brief smell of hot glue, that was gone in another few minutes. I then left the room, with the door closed. I have to say, I did not expect to see that the PM2.5 particulates were elevated over an hour, and the > PM0.3 category was above baseline for over four hours!  You might also suspect the sensor itself got contaminated and was slow to clear, but It went up much higher, and also down to baseline within minutes using a different source ("Wizard Stick" fog/smoke generator) and with the room door open.  Especially in the current times we live in, it goes to show how at least my intuition and assumptions about how long fine particulates in the air stay around, might not be reliable. 

[JBeale]

I was curious, so the following day I repeated the experiment, except that I opened the window and the door to the room after 20 minutes.  It's sort of obvious that the smoke will go away faster with the window open, but it's fun that this little device can actually measure just how much faster.  So I guess this could tell you something about how efficient a ventilation system is.  I know home and industrial HVAC testing systems exist, but with this thing under $20 it may be just about the cheapest way to do it. The initial tall spike was my first test using a puff of "fog juice" vapor/smoke right next to the device.

[DrG]

Nice.

I have previously used a HPMA115S0 (I see that they have since begun to market an updated product) and am currently playing around with a Panasonic SN-GCJA5 see this thread if you have any thoughts on something I am struggling with https://www.eevblog.com/forum/projects/panasonic-sn-gcja5-particulate-matter-sensor-output/msg3088957/#msg3088957 I see that Adafruit has a product with the one that you have and there are others.

I like them, BUT, I have to remind myself about validating sensors. I have always liked investigating sensors of all kinds. For example, I thought it was quite cool to use your own cheap humidity sensors in DIY projects. Gaining more experience I started wondering just how accurate these sensors were and for how long (they frequently advertise some really nice sounding accuracy values). Then I started to validate them using salt standards and I got a completely different perspective and learned about calibration and so on.

So, for particulate matter sensors like these, what do you do? I don't have any good answers other than find a laboratory grade instrument and a laboratory standard concentration of some kind to evaluate these sensors. While you can do this pretty economically with temperature sensors  - buying a traceable high accuracy thermometer to use as a standard, it is not so easy for a particulate matter sensor as far as I can see.

I have gone pretty deeply into some of the literature and there is a lot of information about this e.g., https://ieeexplore.ieee.org/document/6240545 but try finding hard core documentation for what exactly the Chinese cigarette smoke model is, for example.

I guess what I am saying is that I keep in mind their limitations. On the other hand, there are well established issues with breathing a lot of small particles for long periods of time. They are a legitimate health hazard because they defeat your built in defenses and find there way into places that you don't want them (like your blood supply). So, I do think these sensor do, potentially, have their uses in ventilation assessment and indoor air quality and the workplace and so on. I just wish their was a way for people like me to do some validation on them.

[JBeale]

It's a good question about how to do validation as an amateur. The professional sensors are pretty expensive. So far, all I have is the qualitative impression that it is really measuring something, in that I can look at the data and see when I've used a soldering iron or glue gun in the room.  There have been some quantitative tests done by others, like https://www.nature.com/articles/s41598-019-43716-3  but I don't expect every unit to be closely matched or individually calibrated, at this price.

[DrG]

It's a good question about how to do validation as an amateur. The professional sensors are pretty expensive. So far, all I have is the qualitative impression that it is really measuring something, in that I can look at the data and see when I've used a soldering iron or glue gun in the room.  There have been some quantitative tests done by others, like https://www.nature.com/articles/s41598-019-43716-3  but I don't expect every unit to be closely matched or individually calibrated, at this price.

I concur. The two sensors that I have investigated show systematic changes as I would expect and that is, at least, something.

Thanks for the article link, it was interesting. Although I am more interested in indoor air quality for the sensor, I like the idea of correlating outside air quality as measured by the sensors and compared with reference stations. I have used something like that to check on barometric pressure sensors. It's going to the laboratory grade sensor or at least as close as you can get.

The one that I am working with now (SN-GCJA5) has a micro fan and some kind of software compensation - of course none of that is explained well in the data sheets.

[JBeale]

I assumed all these sort of sensors use an internal fan, at least the PMS5003 does have one. It is pretty quiet and I can barely hear it. You can just feel a tiny draft of air with your hand near it.  I've seen some "citizen science" air quality monitoring networks that apparently use these kind of sensors with outdoor air. If you've got a permanent outdoor installation, I do not know how you avoid things like tiny insects and spider webs gumming up the interior air path, without using such a fine filter that it also affects particulates.  At some times of year around my place, it seems like spider webs are everywhere.

[Habropoda]

Purpleair units each use a pair of the PMS5003 sensors.  On the Purpleair map you can choose to apply two different correction algorithms resulting from studies comparing the PMS5003 with reference stations.

I use ordinary window screen material on my unit to block insects and it seems to make no difference to the readings.

Here is what Purpleair has to say about the different methods used:
"PurpleAir sensors use a laser particle counter to count the number of airborne particles in the air. That count is used to calculate a mass concentration, assuming an average particle density in an algorithm developed by the laser counter manufacturer, Plantower. An average density must be used because not all PM of a particular size is made of the same stuff. For instance, PM2.5 from wildfire smoke will have a different density than PM2.5 from dust blowing off a gravel pit. This means that mass concentration reported by a PurpleAir sensor can vary depending on the specific composition of PM for a given area thus making the sensors appear to "read high." So far, two different research groups have completed studies for their areas and created conversion factors specific to the composition of particulates in their air: AQ&U and LRAPA."

"Federal reference sensors typically measure mass concentration of PM by drawing air through a filter and weighing the filter. This method is expensive, difficult to install, requires a specialist to maintain the sensor, and reports on an hourly scale. Because of this, many cities have a limited number of these sensors (or none at all) and it's not feasible for the general public to have their own."

[JBeale]

In case of interest, what my PMS5003 says today about my indoor air, just south of Portland OR.  PM2.5 up to 300, like I've got a few glue guns running in a small room continually.  Outside, above 900 at one point.

[DrG]

In case of interest, what my PMS5003 says today about my indoor air, just south of Portland OR.  PM2.5 up to 300, like I've got a few glue guns running in a small room continually.  Outside, above 900 at one point.

That is, sadly, believable.

[JBeale]

FWIW, apparently this PMS5003 sensor (also used in the "PurpleAir" monitor) yields numbers around 2x higher than the official EPA PM2.5 values for woodsmoke particulates specifically, according to one LRAPA / U.Oregon comparison study: https://www.lrapa.org/DocumentCenter/View/4147/PurpleAir-Correction-Summary

With that correction factor, my outdoor numbers are a much better match to local PM2.5 values shown in online AQI maps.  So my indicated 900 is more like a true 450 (still in the "hazardous" range, outside the usual 0-300 scale).

[DrG]

FWIW, apparently this PMS5003 sensor (also used in the "PurpleAir" monitor) yields numbers around 2x higher than the official EPA PM2.5 values for woodsmoke particulates specifically, according to one LRAPA / U.Oregon comparison study: https://www.lrapa.org/DocumentCenter/View/4147/PurpleAir-Correction-Summary

With that correction factor, my outdoor numbers are a much better match to local PM2.5 values shown in online AQI maps.  So my indicated 900 is more like a true 450 (still in the "hazardous" range, outside the usual 0-300 scale).

Tanks, good to know.

There is the unfortunate lack of an easily available reference standards to validate these kinds of sensors and I am glad to see it being done (as per your link). Personally, I looked for a very long time to find even a formal definition of the "Chinese cigarette smoke model" and the like. I just don't have access to professional level particle counters (like the one used here, for example, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4734356/ ). Even if I did, how would I get a test standard? So, comparisons with published levels (assuming legitimate equipment is being used) as you are doing seems reasonable - especially if you are close to a test site.

[thm_w]

Some are accurate some older ones may not be: https://aqicn.org/sensor/pms5003-7003/

IMO 450 and 900 ug to me are really not so different in terms of a reading, in both cases, stay inside as much as possible. But yeah, if its far off from local monitoring stations, then that is good to know.
I don't recall mine being that far off, but haven't checked when it was that high of a reading yet.

edit: checked one of mine, its reading 255/190 and the closest station some km's away is showing 225/120 PM2.5/PM10, so close enough. I added the 1um + 2.5um together to get a total PM2.5 reading, not sure if thats typically done.

[bifferos]

I don't know if it helps but you can hire a Fluke 985 or 983 from these guys:
https://adastrahire.co.uk/browse/particle_counters, so we're looking at about £140 total for the 983.   That may be indicative of the rate in your country.
Note:  I haven't used that company and I don't represent them, they were just the first hit.
That may seem like a lot, but it's probably not much higher than the cost to the company of having the calibration done.  Will be interesting to see if they supply a calibration certificate with the hire.

[DrG]

I don't know if it helps but you can hire a Fluke 985 or 983 from these guys:
https://adastrahire.co.uk/browse/particle_counters, so we're looking at about £140 total for the 983.   That may be indicative of the rate in your country.
Note:  I haven't used that company and I don't represent them, they were just the first hit.
That may seem like a lot, but it's probably not much higher than the cost to the company of having the calibration done.  Will be interesting to see if they supply a calibration certificate with the hire.

That's a good point and feasible. But you are still faced with a standard. True, you can gain some information just having them side-by-side in various environments (to the degree that you can get a range). I would do it in a minute, especially, if I knew someone who had one that I could talk into using for me. Still, the rental prices were not terribly outrageous.

What I would really like to find are some standard canisters like you find with CO and gasses. Have never seen them for sale commercially (maybe I have not looked hard enough) but have seen them referred to in scientific publications.

[SiliconWizard]

Another interesting kind of sensors you can use in complement is those that give you CO2 and VOC. I used an AMS one.

Monitoring indoors air quality is often a shocking experience. I think it's an established fact that indoors air quality in large cities is often way worse than outdoors. Exceptions may be in extremely polluted areas such as some chinese cities. But otherwise...

[DrG]

Another interesting kind of sensors you can use in complement is those that give you CO2 and VOC. I used an AMS one.

Monitoring indoors air quality is often a shocking experience. I think it's an established fact that indoors air quality in large cities is often way worse than outdoors. Exceptions may be in extremely polluted areas such as some chinese cities. But otherwise...

I also used that AMS iAQ-Core module. I think it is still available under another name. I don't think it measured CO2 directly. Although it gives a ppm value, it is for CO2 equivalents. I interpreted that (rightly or wrongly) as including most any "greenhouse" gas.

[JBeale]

I have a cheap heated metal-oxide sensor, Sensiron SGP30 that responds to various hydrocarbons (VOCs) in the air, but also generates a calculated CO2-eq value based on sensed H2 concentration. The datasheet does not define what that means, but I presume it is this:

Carbon Dioxide Equivalent (CO2e or CO2eq) means the standard metric measure used by [the UN’s Intergovernmental Panel on Climate Change (IPCC)] [in industry] [for vehicles] to compare the emissions from various Greenhouse Gases (GHGs) on the basis of their global warming potential over a specified timescale in order to express a Carbon Footprint that consists of different GHGs as a single number. https://chancerylaneproject.org/glossary/hgv/

[5U4GB]

I ran into this thread while trying to figure out how to test a PMS5003, I've got one that consistently reads 0 for both PM 2.5 and PM 10.  The next-best thing I've got to a check system, a QingPing WiFi environment monitor using the very similar PMS7003 sensor (although some sources say it's a Grandway PM5500), reads 1-2 ug/m3 sitting next to it.

Is there some easy way to test whether it's working at all?  Would lighting a match near it and blowing the smoke into it be sufficient?  There's a smoke sensor connected to an alarm in the same room so I don't want to overdo things too much, and moving the whole setup into a different room is going to be a pain.

[thm_w]

Is there some easy way to test whether it's working at all?  Would lighting a match near it and blowing the smoke into it be sufficient?  There's a smoke sensor connected to an alarm in the same room so I don't want to overdo things too much, and moving the whole setup into a different room is going to be a pain.

Candle, match, flour, baking, frying, anything dusty being kicked up, will easily make the readings spike way up. You'll see it on your other sensor.

Its rare to see all 0 readings, I think I got that a few times but might be a bug in my code. Usually 1-3ug, in a fairly clean environment.

[5U4GB]

Thanks, that sorted it!  A burning match blown in its direction gave a reading and I never triggered the smoke sensor on the alarm which is a pain to reset when it's screaming at you.  It looks like their response curve reads very low values as zero and only gives a reading when it goes over maybe 5ug/m3, I'm still waiting for it to decay to the point where it gives a minimum reading before dropping to reporting zero.  The PMS5003 itself is buried inside a solar radiation shield and readout is via modbus so I don't have access to the raw data unless I were to disassemble the whole thing.

In any case the thing is damn sensitive, just striking a match a little distance away was enough to give a strong reading, and now I know it'll actually work when needed.

[5U4GB]

After two hours it's now decayed back down to zero in a long-tail curve without any sudden jumps, so it looks like it'll read right down to zero, just with slightly different weighting than the other sensor which is still reporting a reading of 4 ug/m3.

[thm_w]

Yeah it takes a long time for the particles to settle or be filtered out.

The fudging of the readout sounds very similar to what is done on soldering irons or weigh scales. If its close to 0g on the weigh scale, a lot of them just round it to down to 0, which is annoying.

[5U4GB]

It did actually legitimately decay to zero, so the last readings were alternating 1 and 0 for awhile before it settled back to zero, there wasn't any sudden jump corresponding to (say) "anything below 3 is zero".  Speculating wildly, it's possible that the initial always-zero readings straight from the factory might be because it needs some kick to start reading nonzero the first time.

For anyone else stumbling across this thread, if you've got a sensor that appears stuck at zero, try the match-smoke kick to see if it's genuinely stuck or just reporting a legitimate very-low reading.