Zwave Thermostat Inaccurate?

[Nestor]

Yesterday I created a program that relies on my home thermostat temperature and it seemed to fail. The value provided by the thermostat is not what is being displayed in the Main devices tab of the EISY or even the physical display on the thermostat. 

For instance, at the time of this post, "Home Thermostat" reads 68.0 deg F and the physical display is 20 deg C (ie. 68 deg F)

I put together a test program to see what value the EISY is working with by simply setting the home thermostat temp to a variable. Result: 65 deg F. 

As further verification, I set the variable to the heat setpoint value and got 68.0 deg F, which is correct. 

So it seems my thermostat is off by 3 deg, but only when being used "behind the scenes" as a value / variable in programs. 

Any ideas?

[larryllix]

After using many different brands of thermostats over the years I have found they alldisplay a different temperature than the actual.

What I have found was that instead of using a different register (memory location) for the calculated /compensated temperature they use the register that is being displayed on the screen.

For a thermostat to have some predictive capability and not undershoot and overshoot, in the mechanical day they would have a small heating element in series with the furnace solenoid, to predict and compensate the time delay in sensing the air temperature that SHOULD be the result of the furnace run. If the thermostat waited until the bimetal element was actually warmed up to the air temperature the furnace would run way too long and overshoot the setting by several degrees. The small heating element tried to "predict" what was coming, in order to prevent the wave of heat that would result without it.

In modern / electronic thermostat days, they attempt to accomplish this "predictive" task arithmetically, by adding a few degrees to the actual sensed temperature and you see it on the face of the thermostat. You may find this difference goes away when it is not calling for heat or cool. You may also find this difference may lessen if you adjust the on/off differential to allow a larger temperature swing setting.

BTW: Thermostats with WiFi or other signal transmitters produce a lot of heat and they manufacturers attempt to compensate this by fudging the calibration of the sensing. This makes a wildly regulating thermostat that becomes very dependent on room air currents. Many are not designed well enough to be good thermostat temperature regulators.

Edited June 2 by larryllix

[Guy Lavoie]

I've had good results with the Venstar T7900 thermostat. The heating element that larrylix is referring to in mechanical thermostats was the "anticipator" winding. This was when the thermostat used a bimetallic coil to sense the temperature and a mercury bulb to switch the furnace on and off. The weight of the bulb made it difficult to obtain a narrow deadband, so the anticipator winding heated inside the thermostat a bit when the furnace was running, making the bimetallic strip come closer to tripping when the setpoint was reached, and narrowing the deadband.

Electronic thermostats don't have this problem, and can be made to have very narrow deadbands, which can also lead to inefficient short cycling of the heating/cooling equipment. To even things out some will have adaptive software, and/or settings to specify minimum run times to avoid tripping off after only a short run time due to a random air current or other transient condition. Some "smart" thermostats will also learn how far past the setpoint the temperature will get after stopping the equipment and then adaptively stop it earlier to avoid overshoots and keep the air temperature closer to the desired setpoint. 

[Nestor]

Are we saying the parameter displayed in the node in the Main tab of the admin console isn’t the same parameter used in a program?

In the example above, the Home thermostat parameter displays 68 deg but when I use the same parameter to set a variable, (eg. Int_Home_temperature), the variable reads 65 deg. 

[Guy Lavoie]

It might help if you'd specify the make and model of your thermostat, and a screen shot of the device as displayed in the admin console.

[larryllix]

On my ecobee thermostats (I have owned three now) there are internal records of thermostat sensors every 15 minutes. These spreadsheets records can be downloaded from the thermostat and show the actual sensor reading in 0.1c degrees, whereas the display and the api sending values are only 0.5c degree resolution. The thermostat face displays the compensated / anticipator resultant temperature the thermostat is regulating the HVAC equipment to. ecobee stats offer a settable on/off differential of down to 0.3c degrees. ecobee stats also insulate the sensor for the electonics heat extremely well with distance, air diversion channels, insulation, and heat reflectors

I have had several Honeywell programmable stats and they do the same thing, displaying the compensated temperature and not the sensor actual temperature. The evidence there was to turn the stat to lower heat setpoint and watch the displayed reading drop by 2-3c degrees. Once you consider the 0.5c "looseness" before the stat contacts will close the temperature swings may exceed 4c degrees or about 7-8F degrees in real applications.

I have owned a few other top tech stats (in their era) and they did the same thing with readings that didn't reflect the actual sensor temperatures. Venstar T78/900?  was another that did the same thing.

Nest stats have the same problem. I find homes with Nest stats have waves of heat and periods of cold at times that I cannot tolerate comfortably.

The 2441TH Insteon stats do not have this anticipation / prediction so the lowest on/off differential you can get is about 3c degrees plus the time delay for the warm air to enter the casing and change the sensor reading. The resultant temperature swings are way too large for many (myself included) to tolerate. However, IRC, the readings you get via protocol, are the readings you also see on the stat front.

Edited June 3 by larryllix

[Ross]

I use a temperature sensor in another part of the room and programs to adjust the thermostat settings based on the sensor.  This method ignores what the thermostat "thinks" is the temperature.