AirGradient Forum

How Accurate is the Sensirion SGP41 TVOC Sensor?

AirGradient Blog
#1

VOCs (volatile organic compounds) are a large group of organic chemicals that easily evaporate into the air at room temperature. They have a wide range of natural and synthetic sources and can significantly impact indoor and outdoor air quality. Because some VOCs can adversely affect human health and the environment, they are often discussed in the context of air pollution and health.

This is a companion discussion topic for the original entry at https://www.airgradient.com/blog/accuracy-sensirion-sgp41/
#2

What is the approximate sensitivity for benzene? I don’t need the absolute values, but I might need to trigger an automation for as low 1-2 µg/m³. Is it expected that such low concentration is reflected in the reported VOC value in any way?

#3
#4

THis sensor isn’t specific to benzene and the values you get out of it is just an Index value that indicates change over time. So it isn’t giving any reliable measure of a concentration for any VOC, much less specific for benzene.

#5

I understand it is not specific. My question is if it is sensitive enough to change the output value if the benzene concentration changes from a long time average of zeroish to approx. 1-2 µg/m³.
(I understand that if the output changes that can be because of the change in concentration of any VOC, so it doesn’t mean that benzene has changed. In my use case that is perfectly fine, I am OK to have any ammount of false positives)

#6

I don’t have that information, and we don’t have a reference instrument for Benzene so we can’t test this. But it awould be interesting to know, so let us know if you find something out.

#7

My use case is that we have an oil refinery a couple of kilometres from our home, and from time to time very bad odours settle on the neighbourhood. The exact composition of the gas is unknown, but many people think it is mostly benzene.

I am thinking about a system which would stop the home ventilation when bad gasses come, for that basically I would need a change in the TVOC a bit before humans can sense the odours.

Just deployed the device today, so I am waiting now for the next occurrence and will let you know. (the DIY version, and I am very satisfied with the packaging, instructions, overall build quality, deployment workflow, so basically with the whole consumer workflow).

#8

Thanks for sharing! Really interesting article!

I was a bit surprised that temperature didn’t influence TVOC readings. If nothing else, I would have guessed offgassing would impact measurements. But perhaps measurements were done in a suitable location, where this wasn’t a concern :slight_smile:

Furthermore, incorporating humidity into the calibration formula does not enhance the sensor’s accuracy. Therefore, the impact of humidity on the sensor’s performance is not significant under the tested conditions.

Kind of surprising, as Sensiron SGP41 offer humidity compensation - so they must think it matters. But perhaps the relationship is just a bit more complex and wasn’t captured well by the linear regression model.

The sensor is limited to tracking relative changes in VOC levels, meaning it can detect when VOC concentrations in the environment increase or decrease but cannot provide precise absolute values.

Also my conclusion based on experience, from way less rigorous testing though! It is mentally quite challenging to try to interpret VOC indexes going up and down though. Time permitting (I’m unfortunately not working with air quality professionally), I hope to explorer using the relative measurements a bit more. A rough idea I have is to try to detect the events, measure the slopes, and do something fun of that. For example try to estimate TVOC reduction capability. In my case mainly to better be able to reason about the effectivness of air purifiers I have - and perhaps when they are due for filter exchange / maintenance.

2 Likes
#9

It’s also possible that the limited humidity range we tested didn’t fully capture this relationship. It would be interesting to repeat the test across a wider range of environmental conditions!

Perhaps we need to highlight this statement a bit better:

Therefore, the impact of humidity on the sensor’s performance is not significant under the tested conditions.

Because I’m sure Sensirion has a reason behind providing humidity compensation. Perhaps this could be an interesting follow-up article.

1 Like
#10

Good point! It could still mean that a linear regression model would struggle, if error changes based on range, by violation of homoscedasticity (I don’t know how much it matters though) [1].

Pity that Sensiron do not share their humidity compensation logic (afaik). My first thought, but perhaps completely wrong, was that compensation might benefit from multiple linear regression models fitted on ranges where error is roughly the same. So some kind of clustering and then linear regression setup.

If anyone has a bit of time and spare air gradients lying around, perhaps a fun simple experiment would be to run (for a day or so) one or two with stock firmware (which doesn’t use humidity compensation based on my understanding of the code), and one with Supply real RH/T when calling measureRawSignals by petercrona · Pull Request #361 · airgradienthq/arduino · GitHub . So a sensor colocation experiment. One extra unit to get data on sensor-to-sensor variability, to reduce risk that humidity compensation on/off does nothing, but looks different as all sensors vary much. EDIT: perhaps 2 with humidity compensation. So the target variable can be if sensor-to-sensor variability is reduced (precision?). Otherwise tricky to say much more than that humidity compensation does nothing or something.

Also, if anyone has ideas/advice on how I could fairly easily test it myself with just one air gradient, I’d love to hear. It feels like humidity compensation helped on my unit, but I realized I have no way to say if it actually did, as I have neither reference sensors or multiple units.

[1] Regression Model Assumptions.