Testing the new kalman-combinator with 3 temperature sensors

I have 3 ds18b20 temperature sensors on different sides of the electricity cabinet outside. When the sun shines one or two of the sensors show temperatures higher then the actual ambient temperature. At least one will always be in the shade, but not always the same. The one in the shade will show the correct value.

I wanted to give the new Kalman filter -based combinator a try, to see what it would report as the actual temperature. Today the sun shone, so here’s the graph:

It doesn’t look too much different from a simple mean of the three sensors. It is smoother though. Smoother even than mean+lowpass filters, I think. I don’t know what the oscillation is about. Maybe I have wrong std_dev values.

Can someone tell me what the std_dev is supposed to be? Deviation of what exactly, and over what time? How can I derive that in home-assistant?

The author of this feature stated in the pull request that

tracking a simple velocity would not be that hard to add

It seems to me like that might improve the result in my case. If Cat-Ion (Valentin Ochs) · GitHub is here, I would very much like to see this improvement :slight_smile: In any case, thanks for this cool component. I had never heard of the Kalman Filter before, and it seems really awesome!