This is disappointing:
The problem is that every one of those devices are within 2 feet of one another and now I am not sure who to believe. The temperature across the Sonoffs and Ecobee are consistent but not so much with humidity. The HTU21s are way off for both readings.
I guess I'm not seeing what you are concerned about?
For example this device.
States that the Relative Humidity precision is +/- 3% which at your 63% average there Would be 1.89% more and 1.89% less than 63 with a total window of 3.78%. Not to mention now that humidity readings with the latest update are only displayed in whole numbers the 62% device very well could be 62.8% and your 64% device could be 64.1% When you account for the 3% all of your devices fall within that range (for the specs of the device posted above)
Same with Temp readings (according to the posted device spec) it varies from .72 degrees F more or .72 degrees F less for a total window of 1.44 degrees F. Your highest reading is 77.40 and the lowest at 76.16 a difference of 1.24 degrees F.
You are all within spec of the stated device above?
What do the published specs for these devices say re: temp/RH accuracy?
I believe OP are comparing devices of different brands.
The HTU21 sensors have both 2 figures way off compared to Sonoff and Ecobee while Sonoff and Ecobee showed different humidity.
I have a Ecobee thermostat controlling my HVAC unit. I also have three Ecobee sensors in rooms around the house to monitor temperature uniformity. They seem to be pretty accurate.
I do not know anything about the HTU21 sensor.
Although digital thermometers often show a precision of one or even two decimal points, that does not mean they are accurate to that level. Most digital thermometers are within +/- 1 degree C/ +/- 2 degrees F. That means that two individual temperature sensors (especially from different brands) can vary by as much as 4 degrees. Since all your reading fall between 76 - 80 degrees F, you are within that stated range.
Hubitat has a "calibration" feature for temperature sensors. If you have an accurate reference temperature, you can enter that into the device handler and Hubitat will calculate the offset to match. Of course, a two point calibration would be better, but even an offset should get it pretty close over normal room temperature variation. If you keep the house at different temperatures summer vs winter and are not satisfied with the results, you can always recalibrate when the seasons change.
So last night, I decided to calibrate all of the other sensors to the Ecobee; right or wrong. I used the built-in calibration feature for the Sonoffs but didn't do anything about the HTU21s. When I woke up, the HTU21s had changed significantly and more closely mirrored the Ecobee in both Temperature & Humidity so I used TempOffset and HumOffset in Tasmota to get them to match the Ecobee.
However, all of the Sonoffs' Humidity went up significantly over night. As you can see they are all either 67 or 68. However, the Ecobee is at 61. I will just have to do a Humidity Offset using Node Red and assume the Ecobee is correct. Since I want all of these Temp & Humidity sensors to adjust my HVAC ultimately, that would seem to make the most sense.
Yes the first thing I did was calibrate all my sensors to a single device.
I bought 3 different sensors (Sonos, T&H, Aqara) and did a single driver for them all so I could adjust temp/humidity offsets as needed so they all now agree with each other.
But even sensors from the same manufacturer could not always agree.
As @waynespringer79 initially pointed out, all temp/humidity sensors have a reported range of accuracy. IMHO it’s always worth checking that first so one can decide if the sensor has an acceptable level of accuracy for the intended use case.
For most sensors that work with home automation hubs (ie not certified to some pre-calibrated standard), that range is almost always plus-minus a few degrees F, or a few %RH.
I have a number of sensors that are +/-5% on RH... Meaning versus another device of similar accuracy they could be as much as 10% apart (5% each way on each device) and still be fully in spec.
Sometimes you can offset that with calibration, sometimes you can't. Depends on the repeatability of the reading, too. If a device isn't consistently shifted, and drifts around within it's spec, then you will always have an offset versus other devices as it moves around.
The way I heat and cool my home (steam radiators with a single thermostat controlling the boiler + mini-split a/c’s in each room), I haven’t yet found a need to average temps across multiple sensors in the same room.
If I were going to, I’d prob get one (dumb) sensor that comes pre-calibrated to a certified spec, then calibrate my hub’s sensors to that one.
Or just pick one hub-connected sensor, calibrate to that and move on 
.
Re-calibrating periodically would probably be needed, to handle that inevitable drift.
I am experimenting with implementing @aaiyar's method of using Dew Point instead of Temperature to control HVAC so as long as all of other sensors match the Ecobee, I think I can make it work. All of the devices would be wrong but they would be basically wrong together.
If i get to the end and it doesn't work like I hoped, it will be REALLY easy to delete my Node Red flow and go back to what I am doing now. 
I’m also about to try out the dew-point method for thermostat mode control (though with only one temp/humidity sensor per room), let us know how it’s working out 
.
A method I use to reduce the range of accuracy from the devices is to average the sensors in a particular room, then again using it for the entire house using this app. There is an app for temp and an app for humidity in this thread.
A man with one temperature sensor knows what temperature it is. A man with two temperature sensors does not know what the temperature is...
A man with 9 temperature sensors will drive himself crazy.
Seriously, 2 feet apart is not good for comparison. The best I've come up with after testing many product temperatures is a simple cardboard box situated away from any draft, heat or sunlight. Let sit for about 2 hours and the devices will be as physically the same temperature as you can reasonably expect.
As for the readings, I use a meat thermometer, Test it in a misture of ice and water ( 0 °C) and boiling water (about 98°C). Then use that as your "standard"
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.
Download the Hubitat app
