cross-posted from: https://lemmy.ml/post/33469454
(Cross-posting this across several communities in hopes of getting some discussion.)
I’m currently building some indoor climate sensors for my home. My idea is to have temperature, humidity, noise, light, VOC and CO2 readings at a relatively high frequency reporting to my MQTT server.
I am currently setting up some different temperature sensors, and I want to calibrate them (hopefully just a linear offset) and evaluate them on some metrics, such as sensor-to-sensor consistency and accuracy.
To calibrate and evaluate the accuracy, I would need a source of truth, and ideally I would also be able to cycle it through a range of realistic values for the given metric.
What are your strategies to tackling these things? Do you assume the sensors are already well-calibrated and don’t bother with this? Do you have a dedicated reader for any sensor value you would want to calibrate?
From the various BMP180, BME280 and DHT11 etc I used over time, I’d say they’re not calibrated very accurately. I had some next to each other and they all reported slightly different values within a range if something like 2°C… Idk. The tolerance should be in the datasheet. The BME280 for example should be accurate to 0.5°C. Other sensors aren’t as accurate. And with some other sensor types, the measurements also depend on temperature and other factors.
Could you tell whether they were just offset by some value, or if the measurement response was dependent on temperature as well? I hope it will just be a linear offset and that they are otherwise fairly accurate within their specifiations.
I am about to do a test over about a day to compare a handful of sensors (BMP180, TMP117, LM35, DHT22 and the onboard sensor on the Pico) to my Netatmo station. I don’t really trust those readings either, which is why I am considering some better quality calibrated measurement device to calibrate against.
Sorry, I didn’t do a proper test so I wouldn’t know. I’d guess it’s just an offset. After all they’re made to measure temperature with some accuracy. I guess it also depends on your enclosure, since that can be hit by sun, or trap the heat losses of your electronics which will dissipate heat. And I believe some sensors warm up a bit once you set them to really high frequency measurements. I’m mainly doing smarthome stuff, so it doesn’t really matter whether it’s really 20°C or 21°C in the room.
And I had a bad time using DHT11 and DHT22 outdoors. They’d go into saturation with the humidity measurement and report 100% for a day, and two of them broke after a year and a half or something like that. I’m not sure if that’s the common experience but I don’t buy those anymore.