Almost all low cost air pollution PM sensors use the poorly documented Chinese devices from PlanTower - and are very happy with it - which is great. However, if one wants to design a long term, high quality (as much as one's pocket can afford) public air monitoring stations, are there any better alternatives?
I am developing one such opensource device - YAAPM (Yet Another Air Pollution Monitor) called Breathe2. This is based around a Sensirion SPS30 which is twice the cost of PlanTower units, but is from a reputed sensor manufacturer with guaranteed 8 years continuous operation life. All is great, but PM10 results are abysmal (also observed here in comparison to reference instruments). The kind people at Sensirion have been very responsive on many issues (helped solved all) but refuse to acknowledge this PM10 issue.
Question: has anyone used Sensirion's SPS30 before? Are there any tricks around it to give realistic PM10 results? Are there any good alternatives to PlanTower?
First, I haven't used the SPS30 so can't comment on specifics for those sensors. However, I have many years of experience working with aerosol measurements and I have worked with the Plantower units for several years. The short answer about PM10 is some variation of "tough luck" ... more in detail, there are two main sources of "error" in PM measurements with optical sensors, particularly with "low-cost" ones. The first one is the fact that the sensors need to make assumptions about the density of the material they're sampling. In practice this translate into sensors that were "calibrated" in biomass smoke compare very well to "standard" instruments in wood-smoke dominated environments while not so well on soil or sea-salt dominated environments. The second, and I believe most important, source of error is the inherent noise in the "larger" particles and how that affects the measurements. In a "normal" volume of air (rural, urban, marine ... anywhere except inside a lab) there are many times more particles that are smaller than 1 micron than larger than 5 microns ... BUT because one 10micron particle will weigh 1000 times more than one 1micron particle, one large particle more or less that the sensor sees will change the reading significantly. You can see here for a longer take on low-cost sensors for PM.
Now, coming back to the SPS30 and looking at the AQMD results it seems that the SPS30 cuts the larger particles. It either doesn't see them (the light sensor has a high limit for its response) or they get mechanically removed because of the path that the air follows to get to the sensor. BUT, without opening one I can't say for sure.
As for the second part of your question on alternatives to the Plantowers ... several, the SPS30, Novafit, Alphasense, Honeywell. All work pretty much the same way and if you are looking at PM2.5 and PM1 (and as an air quality scientist, that's where the focus is ... PM10 is definitely old news), then all of them do a decent job.
Hey thanks a lot! Your reply is very interesting and so are your explanations. Here's a 'tear-down' of this sensor, and with fantastic explanations.
As of the PM10, i just received a response from Sensirion which says that they estimate PM4 and PM10 based on PM1 and PM2.5 values. This is because due to the rarity of heavier particles (which you also mentioned) the sampling time required for proper PM10 measurements is greater than 5 minutes whereas they wanted PM2.5 to be sampled every 1 second. As per your notes, everything is an estimate, isn't it?
Now the question arises as to what sort of particle distribution (gaussian, log, smoke specific) is used to estimate PM10 from PM2.5 - it could be different in different geographic locations as well as indoors and outdoors. By the way these sensors are for indoor use (Sensirion email communication) but i must use it for outdoor AQ measurements. As of the calibration, the datasheet and Sensirion responses mention that PM2.5 values are calibrated using fine potassium salts - this could indicate a great improvement over biomass stove calibration method? I don't know how well this relates to traffic pollution?
Since you have worked with many such sensors (thanks for the alternatives list) do you think others also 'estimate' PM10 rather than measure it? The physics and statistics of all these sensors will more or less be the same, right?
Thanks again.
Is this a question? Click here to post it to the Questions page.
Great tear-down!
I see that a lot of love went into the Sensiron design and I mostly agree with the comments ... the issue of the HEPA filter is more dubious in my head. It may be put there for that purpose, I just question its performance.
Nevertheless, back to the PM10 issue. Your question about which size distribution is used to estimate the different PM metrics is key and unfortunately there is no "correct" answer. The calibration info that Sensiron provides ("fine potasium salts") tells me that they use an aerosol generator (nebulizer, or VOAG, or something that takes a salty solution and makes a mist with a controlled concentration of particles) which means that the size distribution of the aerosol is more or less known (and I'd venture that it is close to mono-modal with a peak around 1 or 2 microns).
This is not inherently good or bad but tells me that their calibration factors are not suited for any type of combustion source so if you plan to use a Sensiron unit outdoor you'll first need to do what we do at the beginning and end of every campaign, which is to run the sensors exposed to air similar to what they'll measure during the deployment AND together with a "standard" instrument (and I'm sorry but the DusTrak doesn't qualify in my book), compare the measurements and extract correction factors that are appropriate to that specific deployment.
Finally, yes all optical sensors estimate PM metrics because PMx is DEFINED in reference to the weight of a filter, not an optical measure. Different sensors like the Grimm 11E or Teledyne's T640 have procedures that when you follow them you can claim that their measurements are "equivalent" to the standard but even those can and will report different concentrations if exposed to different aerosol, even if the amount of mass of particles remains the same.
The physics of the measurement principle of all those low-cost sensors is the same but their code implementation will be different. Some use smoothing functions which causes their response time to increase.
Wow, thanks once again. Its been many months searching for such information from 'experts' i know locally but none were so clear and responsive. Really appreciate this help!
So its clear now - the only way to make desent sense is to place this Sensirion SPS30 near a reference instrument (as close as legally allowed, i have no money to buy one) and obtain correction factors (which will vary for PM type) for all PM values.
PM sensing is so complicated, but necessary. I know of 2 efforts to create opensource PM sensors 1) UC Berkeley team and 2) Federico Pedersini (U of Milan). The latter one is very promising because it modifies a regular Sharp dust sensor to achieve good granulometric distribution - which can help identify the type of PM source i guess. Have you come across other efforts or actually built similar opensource PM devices?
Thanks a lot. subir
Is this a question? Click here to post it to the Questions page.
Reply to this comment...
Log in to comment