Hacker News new | ask | show | jobs
by spladug 2407 days ago
The South Coast (California) Air Quality Management District does some really thorough tests on the quality of the various low-cost sensors: https://www.aqmd.gov/aq-spec/evaluations/summary-pm
1 comments

Anyone want to review this table and suggest a currently commercially available option in the $300 and lower price range? I may have time this afternoon but if anyone does go thorough it I'm interested to know.
Based on my reading of the table, the PurpleAir II sensors seemed to be the best of the sub $300 for PM1.0 (field R^2 of 0.96 to 0.98) and PM2.5 (field R^2 of 0.93 to 0.97). The PM10 readings were not as good (field R^2 of 0.66 to 0.70).

After skimming the rest of the table, it looks like the PurpleAir II sensors might have some of the best field R^2 for PM 2.5 and PM 1.0

That sensor is also available from Adafruit: https://www.adafruit.com/product/3686 I got it hooked up to a RaspberryPi that exposes sensor readings in Prometheus format.
Do you have a PurpleAir II to compare against? I suspect that there will be some extra calibration or signal processing some to make it more accurate that will be missing in the raw sensor.

EDIT: The link we're discussing says this explicitly: "[...] These particle counts are processed by the sensor using a complex algorithm to calculate the PM1.0, PM2.5 & PM10 mass in ug/m3. [...] PurpleAir PA-II uses two identical PMS5003 sensor units attached to each other and placed in the same shelter. [...]"

I don't think you can recommend the PMS5003 as a substitute for the PurpleAir II.

That sensor available at Adafruit uses the same algorithm for calculating PM levels, it's literally the same chip. Maybe using 2 of them yields more accurate results though.

The specs for particulate matter are identical:

https://www2.purpleair.com/collections/air-quality-sensors/p...

https://cdn-shop.adafruit.com/product-files/3686/plantower-p...

The Adafruit sensor is just the sensor, the PurpleAir II is two sensors plus extra logic that processes the readings to give extra accuracy.

You can't expect one sensor to give the same performance as two sensors plus correction logic. If it was that easy, PurpleAir could just put a case on a PMS sensor and be done with it.

I agree, looking at devices around the sub-$300 price point that were lab tested, PA-II has good performance and seems to be readily available.
If you just want a sensor rather than a full retail device I'd recommend the SPS30. I have one and it's very easy to work with and per that table, is very accurate, while not needing any extra software correction. It's also cheaper than that table suggests: you can buy it for ~$50USD from Mouser.
I have one of these and it's easy to hook it into your home automation stuff as it presents a CIFS share and it's simply a matter of attaching and reading a text file.

However, the damn thing turns on it's front display and backlight randomly at all hours of the day and night. If you are a sensitive sleeper this thing definitely can't be on your list of items. I contacted their support and was ignored. Generally wouldn't recommend.

you can try changing the power saving mode (under 'performance'). it lets you adjust when the screen auto turns off/on
>*Field R2

Under the table is definition of R^2

>The coefficient of determination (R2) is a statistical parameter measuring the degree of relation between two variables. ....

The more - the better. Range [0; 1)

More device in your area === better quality of measurement

Isn't the table sorted alphabetically?
True

My bad