High humidity destroys PM sensor reliability

Last week I decided to put the SDS/SHT unit inside our (old wooden) Stevenson screen, but last night I got ridiculously high PM readings:

and the unit started becoming very spotty as you can see.
Apparently it just can not handle really high humidity values:
I had not seen this before with this (fairly new) sensor-kit, but that was before it was inside the Stevenson (and this particular sensor-kit had not seen such high humidity values yet).

Anyway, I took it out (and indoors) a few hours ago, and the PM values dropped to normal again. But honestly, this is really bad. RH above 80% values are very common here during nights, it’s in a field outdoors, so essentially this means we can’t put the sensors outdoors.

Is there any experience with this by others? Are these misreading related to humidity/condense on electronics, or are they related to the actual SDS sensor not getting it when there’s fog on top of its membrane or something?

1 Like

Yes, this is normal behaviour. With a humidity of 70% and up the particles are absorbing water and the readings become higher. I am currently experimenting with different techniques (calibrating with multiple lineair regression, heating the airflow to reduce humidity (take care of the T/H sensor!) ).

A lot has been written about this. Do soem extensive searching on the internet on the use of low cost sensors and you will find a lot of information.

1 Like

I cannot recall seeing such high PM values in high humidity. Maybe if there is fog but otherwise not. Have you checked official data and/or other sensors near your place? Maybe the PM values are realistic. At my place, there was a pronounced PM increase (“just” up to 35) for several hours. I also suspected a humidity effect but the official PM station data showed that the pollution is real (actually for the whole North of Germany).

I am seeing occasional dropouts in the plots, too. Sometimes it’s a local Wifi issue. Most of the time the data is visible in my local influxdb, hence I believe it’s a communication problem with the remote server or a problem on the remote server itself.

OK, I have placed the sensor-board in a drier place (with more air-flow - and an outlet from a warmer dryer indoor area) and hope it will make the readings more reliable. The sensor inlet hose and SHT31 are still in outdoor air this way, shielded from the ‘heated’ board, so that should work.
Either way, having it inside our Stevenson was not a good idea, it’s just too humid there. It would just break down giving strange readings and drop-outs each time there was a fog of some sort.

I’ve got a cheap solution for high humidity scenario.


HECA was designed for use in Nettigo Air Monitor, but version 2 is also compatible with inline tube like the one used in Sensor.Community. It’s small. BOM is cheap. Parts are widely available.

Aluminium extrusion was used because it reflects IR and make the system more efficient.

In NAM we use it inside enclosure with side inlet.

But it was designed to use with two black caps with hose connectors on both ends. So it can work as inline air conditioning device.

Current form factor is retrofit for NAM 0.3.3+ series. Also 10Ω resistor powered from 5V has too much power for typical use case. Fortunately value can be easily adjusted by changing the resistor.

1 Like

I wasn’t sure that the sensor community supported heating the air input as there isn’t a way of recognising this in the data for the community map. It would be really useful if this could be added to the data as where we live the humidity is usually high.

If anyone wants to get really into the problem I found this on researchgate (PDF) Comparison of a Computational Method for Correcting the Humidity Influence with the Use of a Low-Cost Aerosol Dryer on a SDS011 Low-Cost PM-Sensor.
As far as I can tell it isn’t peer reviewed but has some interesting observations.


I see transistor on schematic, why not to use pwm to reduce power if it needed?

I think there are more heat transfer by conductive, not radiation way and every plastic (may be with foil) be more efficient with heater.

The real problem is water droplets in very high humidity. Many have tried to build an algorithm to compensate for this but none have succeeded. I believe the only solution is a heater.

The foil should be earthed to prevent static which removes particles.

Can it be switched off when humidity is below 70%?

Have we come to a definitive answer on the approach to high humidity and obtaining more accurate PM data?
I still think that Bernd Laquai et al ((PDF) Calibration Method for Particulate Matter Low-Cost Sensors Used in Ambient Air Quality Monitoring and Research) have a good method. The issues with high humidity being 2 fold: that of water droplets and hygroscopic growth on particulates. The article suggests heating the air to remove the water droplets and then monitoring the humidity of the exhaust to determine the amount of water still associated with the particulates. An algorithm can then be used to determine the actual particle size.
As the type of particulate matter varies with location this needs to be calibrated vs a reference system for optimised accuracy. Is this possible to achieve? Or is the community happy with the reduced confidence in the high humidity results? I am considering lobbying my local council to purchase a portable reference system but implementing the Laquai changes would mean a diversion away from the sensor community preferred set up. What do you suggest as the best way forward?