Back to

More info on how measurements are taken

Hello, I feel that an expert would find the answer to my question fairly obvious so I’m just going to ask it since I didn’t see what I was looking for from scanning the documentation.

Are measurements taken with the urban board taken as an average of sample over a period of time or are they one one sample? To be more specific and to the point, Is the noise sensor running a function like max or average on multiple samples before it is reported on or is it just one sample?


Thanks for posting!

Measurements are taken as a single reading in what we could call a reading event. These reading events happen every reading-interval amount of seconds, by default 60s, except for the PM, which is 300s. If the sensor needs time to warm-up, stabilise or something like that, then the we allow for that time, but we only take one reading and report it. In the case of the noise, samples are taken for the necessary buffer to be filled (256 samples at 44,1kHz which leaves aprox 20ms) and we make FFT and A-weighting, for later RMS calculation, which is what we report to the platform.

At the moment, there is no implementation to be able to read several samples every time the reading event happens, so there is no way to average samples that are taken unless you use separate ones from different reading events.

Does it make sense? We can clarify this on the documentation for future reference.


Thanks! This completely answers my question.

1 Like

Hey @oscgonfer

Does taking PM readings more frequently than every 300s/5mins degrade the quality of the reading?


Hi Cyrille

As I understand it; The main reason readings from PM sensor are taken every 300 seconds is to reduce power consumption. The device contains a semiconductor laser and a fan and likely the latter is power hungry.
I cannot comment as to whether reading it more frequently will degrade reading accuracy or not.

However; here’s a thing. Most people have the SCK 2.1 that has ONE pm sensor. But the larger SCS V3 has two PM sensors. They are read alternately. The power feeding to the units is controlled. When it’s time to take a reading; power is applied and the reading is taken after 15 seconds. So, using this method readings can be more frequent. But it’s about a one in 10 duty cycle, saving a lot of power affecting battery driven instances of the kit.
I have noticed that the readings from each sensor show differences. It’s possible to average the readings but this would need to be done externally. The differences/variance underline the fact these sensors are the ‘inexpensive’ variety; not laboratory instruments. The are ‘indicative’ of dust pollution levels; not precise measurements. The differences seem to be of the order 10%. (2ppm out of 20 ppm); so still useful.



@cyrille.mdc, what @bryn.parrott says about power consumption is right.

In terms of data processing, just keep in mind we have a data instance available for custom data processing: Custom data processing - Smart Citizen Docs

Taking the readings more frequently no, it doesn’t degrade it. It only needs to be taken into account that the readings happen in a one-shot mode so doing it more frequently would need to ensure that the stabilisation period is still covered. This stabilisation period is 15s, and it only needs to happen once if the sensor is continously operated, but if not, it needs to happen every time.

Important note for new batches of PM sensors: as reported PM Sensor always reading 0.0 - #24 by victor we are still working on it for the new sensors, and we think that the 15s stabilisation period is not going to be enough.