Usage Notes

StanJ wrote up an amazing analysis report of using the PM2.5 sensor in their lab, and we think its helpful for others to understand what and how the sensor works and what to expect from it! We've duplicated it here as well:

I've read quite a lot on the PlanTower sensors, although I'm nothing like an expert :-). The CF readings are 'Calibration Factory' and aren't useful; the 'Environmental' or 'Ambient' concentration readings are the data you want for air quality measurements. I'm using the PMS5003 for a continuous check on cleanroom quality, so I only use the raw Particle Counts as that's the measurement specified in ISO 14644-1 .

As Solaria123 noted in viewtopic.php?f=19&t=135496, the sensor estimates particles > 2.5um and doesn't (or can't) measure them. The article at ResearchGate showed that a concentration composed solely of larger particles wasn't seen by the sensor. For our cleanroom use that's OK as the HEPA filters are more efficient as the particle size increases. For non-filtered air it's a bit more of concern as the different particle sizes are composed of different pollutants, so you might be missing a pollutant if it's composed primarily of larger particles like pollen.

One amusing note in the translated PlanTower datasheet is "Only the consistency among the PM sensors of PLANTOWER is promised and ensured. And the sensor should not be checked with any third party equipment." Several groups including AQICN.org have done exactly that, and we have as well. The PlanTower sensor compares favorably with the readings from our calibrated Beckman Particle Counter, although the 30-50% uncertainty on the PlanTower 0.3 and 0.5 um bins means you can't get an exact comparison. We're only using the sensor for a rough check on current air quality, not to verify compliance with ISO 14644.

A frustrating artifact of the PlanTower sensor is the sampling rate versus data output. With small change between readings the sensor only updates the counts every 2.3 seconds, although it outputs data every second. That means it may duplicate over half of the data, with no way to verify whether any reading is a duplicate. For a normal home or outdoor setting you could simply discard any reading when the checksum is identical to the previous data, as you're highly unlikely to have two successive samples with the same values. In a cleanroom we're looking at very low particle counts, and two successive samples might well be identical. The only way I could get around that is by throwing away 2 of every 3 data packets to insure I'm getting real counts, which increases the total sample time. I add the results from 100 unique 0.1 liter samples to get a reading of particles in 10 liters of air for my measurement, which means 300 samples with 2/3rds of the data thrown away.

Download: file
    amb=[003a 005c 0061] raw=[386a 1160 0325 004c 000b 0001] csum=0542
    amb=[003b 005d 0063] raw=[38cd 1175 033c 0054 000e 0004] csum=05ea
    amb=[003c 0060 0065] raw=[398a 11ba 033c 0054 000a 0003] csum=05f4
    amb=[003c 0060 0066] raw=[3a8c 120f 0340 0050 000d 0003] csum=0555
    amb=[003d 0060 0066] raw=[3b04 122e 0333 0050 000d 0003] csum=04e1
    amb=[003c 005e 0064] raw=[3b04 122a 0339 0056 000b 0003] csum=04dc
    amb=[003c 005e 0064] raw=[3b04 122a 0339 0056 000b 0003] csum=04dc duplicate
    amb=[003c 005e 0064] raw=[3b04 122a 0339 0056 000b 0003] csum=04dc duplicate
    amb=[003c 005c 0062] raw=[3b22 1232 0330 004b 000a 0003] csum=04e2
    amb=[003c 005c 0062] raw=[3b22 1232 0330 004b 000a 0003] csum=04e2 duplicate
    amb=[003c 005c 0062] raw=[3b22 1232 0330 004b 000a 0003] csum=04e2 duplicate
    amb=[003b 0059 005f] raw=[3a7a 1211 030e 0043 000a 0003] csum=04de
    amb=[003a 0058 005e] raw=[3a7a 1211 030e 0043 000a 0003] csum=04d8
    amb=[003a 0058 005e] raw=[3a7a 1211 030e 0043 000a 0003] csum=04d8 duplicate
    amb=[003a 0058 005e] raw=[3a35 11fa 030c 003b 0009 0003] csum=056e

What you're seeing above is the 1 second data window sliding along the (typical) 2.3 second sampling window. When the data changes significantly between samples the sensor shortens the sample window to 200-800ms, which may be why the first 6 data points show unique numbers (faster sampling rate).

The readings above are in my home, and I smoke so the particle counts vary wildly about 1000:1 over time with a decent quality air filter. When I'm home I run the air handler fan continuously to level out the temperature over the house, and when I'm away I let the fan cycle with the AC or heat. You can see the difference below in how rapidly the particle counts fall off with continuous filtering. The rapid fall off continuous curve is [sleeping], and the slow fall off is cycling [away from home]. Data points are every 30 minutes.

This guide was first published on Dec 27, 2017. It was last updated on Dec 27, 2017. This page (Usage Notes) was last updated on Nov 12, 2019.