For what it's worth, I knocked my system together with the digital sensors etc quite some time ago.
I spent too much time thinking "well, that's ace", as opposed to actually testing the product.
At some point I purchased some cheap-■■■ digital thermometers and ran some comparisons. The (smart) temps were way outside of reality.
Common sense kicked in and I firstly realised the rooms had natural hot and cold spots. I therefore relocated the sensors to more suitable locations, Eg near the sofa etc.
Results were better, but still a bit wank.
I then realised that (obviously) the readings were only as good as the airflow over the sensor. Plus the sensor needed to be kept away from the nodemcu's as these little sods were creating heat of their own.
Eventually after a few prototypes, I found that by mounting the chips on one isolated side of a double/single backbox and keeping the sensor seperate in the single..... then periodically using (2) 5v fans and some vents to draw air over the dht22's (which was mounted inside a tube, hidden in the single side of the back box) kicking it back out through the 2nd vent..... Created perfect results. With a reaction time of a matter of minutes, matching with the results from my visual sensors.
Result =)
