Ability to filter on sensor type and / or idenfity faulty sensors

Looking at the sensor.community map in my local area there is always one or two sensors (which I didn’t make) which are not showing correct values. They are often using Plantower 5003 so even when they are working they show different values from my SDS011 based ones. It would be great to be able to do one or both of the following:

  1. Filter by sensor type so for instance I simply would not see Plantower based sensors.

  2. Allow us to email the sensor owner via sensor.community anonymously. So just as a sensor host gets an email when they don’t send data for over 24 hours, they could get an email when someone notices that the sensor is not working properly. Even better of course would be to check neighbouring sensors once a week or so and identify the ones that are not working properly.

You can see an example of how this has already been done using the publicly available data:

Scroll down to the ‘Neighbor Sensors’ table and you will see that one of my neighbour’s device is faulty.

Thanks for your consideration of my ideas.

1 Like

Hi @Graham21,
This is a great idea! I also sometimes see a sensor there, where normally my one is, although I’m the only sensor in the area. I always refresh the page and then it suddenly disappears…

1 Like

An interesting request.
What criteria would you use to determine a sensor isn’t working correctly?

I mean beyond the formula in the attachment.

I guess that would depend on whether there were other sensors nearby or not. I look at the map all the time and there is a sensor near me that’s clearly not working because the colour is always out of step with the ones around nearby. If every sensor is red or orange or purple in a 10km radius and you see that one is green then there is something wrong. I admit it’s more difficult the other way around, a purple sensor in a sea of green might be a local fire, but if it’s always like that then the sensor is faulty.

We had a University initiative here which led to a well known locally based distributor shipping out pMS5003’s which people then use to submit to Luftdaten. You could argue that they are more respresentative (not sure they are) but the point is if 99% of the sensors on the map are SDS011, I’d like to be able to filter out the ones that aren’t SDS011. By the way the local sensor that always shows higher readings is a pMS 5003.

Thank you for getting back to me. Although it is a more than reasonable assumption, I am not certain that one sensor constantly reading lower values necessarily means a faulty device. It could be that the individual has set it up in a location that is less than ideal, or is using it in a different way than intended. I have a similar situation with one of the sensors near me where we are all using SDS011s. As it is lower I usually ignore the outlier in any analysis.

Do you calibrate your SDS011s against each other? Currently, I have three individual SDS011s placed in the same location (one is heated). Even at humidities below 70% there is a variation in the output between all of these sensors that can be as high as 40%. I use the mean of the three and a calibration factor to try and get a more “accurate” reading from the individual sensors.

If you do calibrate your systems against each other would you be willing to share your methodology? I hope to have more devices installed locally and I am trying to decide on the best way forward.

Thank you in advance.

I made a prototyp map with sensor choise in drop list a few years ago and documented it here. Unfortunately it had no success. We can talk about it also.

1 Like

Hi, I usually make them in batches of 8 - 12. I put them in the same location for a week or so and see if any are wildly different. Of course they vary. They are placed at different heights, they might not sample at exactly the same time (I connect them to a 10 port USB hub that allows me to reset them at exactly the same moment - that might help). I understand that they are +/- 10%. I’m not that bothered. Mostly I see that the hexagons are all the same colour on the map, unless some are faulty or someone is burning something. PM2.5 doesn’t vary that much from place to place. I want to use these to show people the affect of burning a bonfire or using a wood burner and I can easily demonstrate that with an SDS011. I just don’t like it when there are PMS5003 always showing a different colour I’m in Sheffield by the way.

There was the site Particulate Matter - Sensorcheck but I don’t know if it still works.


Where I live there is also a sensor that just keeps on showing (much too) high values, and I cannot explain to myself why. Either they are burning something, they smoke 24 hours a day next to the sensor or they have right next to an air-conditioning output… :slight_smile:
Whatever the reason is: it makes no sense.

For example this sensor. I do not understand how this happens. Perhaps a mistake with the sensor, but it’s unlikely.

That sensor appears to have the exact same problem with this one.

I think there is something wrong with these sensors, it is easy to tell by the fact that the pm2.5 and pm10 values are almost the same.

1 Like

Thank you, I have 8 due to arrive soon, so I will set them up as you suggest and run them together with our other three for a week. That should allow some idea of the performance of each one against my “references”. I have noticed that as my systems age they seem to have a higher than +/- 10% difference. I haven’t explored it too much but it might be something to consider as the SDS011s age. I have an Atmotube system which seems to agree with the average of the SDS011s.
My intended use is to determine if the count reduces with time, and (if possible with these sensors) if there are local sources of PM2.5 where the sensors have a sustained reading above the others.

The sensor you show are according to their ID already quite old. Maybe they were not well installed or let in the dust etc. There are quite a lot of reasons why the SDS gives such results if the person does not check the AirRohr from time to time

Normally the PM2.5 & PM10 values differentiate by quite a bit and above there is hardly a difference. There is something wrong definitely.