Calibration_factor attribute for sensors?

[robmarkcole]

Many sensors are reporting measurements of physical factors, e.g. temperature. As noted in the forums, some sensors (e.g. Hue) have a known issue where the sensor values are systematically incorrect. This can be resolved by applying a factor to the sensor value. As this is potentially a common issue, perhaps we should have a calibration_factor attribute for sensors, such that sensor_value = sensor_reading * calibration_factor?

Ref robmarkcole/Hue-sensors-HASS#106

[OttoWinter]

Well the thing with that is that it can get arbitrarily complex. In my experience, most sensors have a constant offset to the real value - especially for temperature and humidity sensors. But you can go all out and make complex formulae that map the sensor value to the "real" value - for example with two fixed points.

Also, I don't think it's really possible with the current architecture to achieve an option for all sensor platforms. I would recommend using the sensor.filter platform (and adding filters there if necessary).

[robmarkcole]

Yes I agree it can get complex. My training is as a physicist, and if we wanted to be really precise you would characterise the calibration_factor as a function of e.g temperature & humidity etc. As you say we have other methods in HA to post process readings

[yottatsa]

From what I know from civil engineering, the temperature in a room depends on height non-linearly, and usually differs for about a 1.5*C / 1m. Moreover, our perception of room temperature is also depends on temperature on different heights, like the temperatures around legs and head.

So instead of calibrating each sensor to be precise I’m placing them on different heights and averaging to the abstract room temperature. This adds both precision and redundancy to the setup. I’m planning to release this sensor.room CC when I’ve got more time.

[robmarkcole]

I think the consensus is to close this issue then. Will leave for timebeing to give others a chance to chime in