Looking for zwave thermostat with 0.1 degree F increments

💬 Lounge
1

TLDR version: I''m trying to find a zwave or wifi thermostat that reports and can work in 0.1 degree F increments. If you know of one I'd appreciate it if you could point me to it. Thanks for that. I guess 0.2 or 0.5 degree increments might also work well enough, but 1-degree increments aren't.

Backstory if you care to know more: I have two buildings that each have hydronic radiant heat (water-filled pex tubes in the concrete slab). After being frustrated by the temperature control I could achieve with either a Watts wifi thermostat designed for this type of system (I hate that device) or a conventional zwave thermostat with 1-degree increments (better but not perfect), I used Ecowitt temperature sensors (report in 0.1 deg F increments) and a zen 16 switch/relay device to control the system in one of the buildings, using Hubitat. It works amazingly well and maintains the correct temperature much better than either of those other thermostats, without excessive cycling of the system (which is what happened with the Watts device). The problem is, now I'm the only person who knows how it works. I'm on a sort of mission to make our various systems more conventional so that if I die or am incapacitated, a service person can understand and work on them for my wife. So... I think a thermostat that reports out and can act on increments less than a 1-degree change would do it, with 0.1 degree F increments mimicking the system I have now, but in a way that the technician from the local service company can deal with.

2

Honeywell T6 Pro Z-wave can do .05, I think the TBZ (If you can find them) does .01

@aaiyar Which one do you have?

3

Honestly, the T6 pro does 1.0 or more, due to Celsius rounding. Enabling Zwave temps makes it worse, it uses an algorithm to try and make F conversion fit the decimal scale better, but then it is even further off for some temps.

Without Zwave temps on, it will never even show 74 on the screen due to C rounding. I'm not sure what funky math is behind Zwave temps, as it will create a F decimal number that shows 74 on the screen, but it is not a real reading.

Without Zwave Temps On:
23 C = 73.4 (thermostat displays 73)
24 C = 75.2 (thermostat displays 75)

" AI Overview

The accuracy of a Honeywell T6 Pro thermostat can be affected by placement, with an advertised accuracy of approximately ±1 °F but user reports indicating potential issues with temperature reporting and readings being off by several degrees"

1 Like
4

I was assuming F not C (As I'm american)

5

Here’s a list of all US zwave thermostats along with their manuals. IDK if what you want exists, but its as good place to start eliminating possibilities.
https://products.z-wavealliance.org/?_zwa_device_type=thermostat-hvac&_zwa_region=us

2 Likes
6

That is exactly what I do because my T6s are so horrible with temp control. I made my own anticipating virtual thermostat, I use Zigbee Temp sensors, and I have a Zigbee relay board at the furnace that controls the zone valves instead of the thermostats. A controller app feeds temp to the virtual thermostats, and reacts to the virtual thermostat state to control the zone valves. This is my third winter on the system, it has not had a glitch and my house maintains a temp within a .5 degree swing.

2 Likes
7

So, after working in calibration labs, where dimensional accuracy is a requirement since the late 80's, the best temperature controls I have seen for a lab were maintaining 68 +/- 0.5 deg F overall temperature with drift of 0.1 deg F/hour. You do not want to know the costs of such a system. It relies on much more than just a thermostat. It relies on creating specific types of air flow. I have yet to see a consumer version of any system that would maintain the accuracy (much less the stability). I would doubt the claims of anyone who sold one at consumer pricing.

Just because something has 0.1 degree resolution doesn't mean anything. Any digitial signal can have infinite resolution. It is effective sensitivity (the smallest change that actually increments a change in display), and overall uncertainty that you would be looking for. Again, that would likely not be something that is affordable as a consumer option. But, if you have ~$100K or more, there are those who design systems (because it requires more than just a thermostat) that actually will maintain the temperatures you mentioned in the OP and not just say that it does.

6 Likes
8

Thanks for the replies. May just need to plow through those manuals to see what has the lowest increment.

As for the 0.1 degree F sensitivity, what I do now with these inexpensive Ecowitt temp sensors DOES work remarkably well and it darn sure didn't cost $100K+. For the entry foyer zone I have Hubitat set to watch for the temp to fall to 66.9. If it stays there or below for 10 minutes (which allows the heat to "rebalance" from normal coming and going through the door) it turns on the relay on the zen 16, which turns on the circulator for that zone. It will usually run for 3 hours or more and then stay off, sometimes for a full day or more, which is perfect for a radiant slab. It turns off when the temp reaches 67.1 and stays there or above for 10 min.

For the garage I have it set to come on if the temp drops to 59.9 and stays there for 30 min or more, which provides a longer period of readjustment because the garage doors are larger and more heat is lost when they open. It turns off at 60.1 and stays there for 10 min.

BTW, I've tried using the in floor sensors but find that method provides less control when outside temperatures vary considerably.

The more I think about it, it might just be easier to write all this up so the tech can understand it and install a tablet for a display. Seems somewhat easier than trying to work with these thermostats.

1 Like
9

I was not saying the sensors would cost that much. A system that actually controls your temperature to within 0.1 degrees does though. As to the sensor, if you believe it is accurate to 0.1 degrees F, then that works for you. From my experience, you are not going to find a sensor that is actually accurate to 0.1 degrees at a consumer price point. It may be good at icepoint (or for real calibration - triple point). It may also repeat. But it isn't going to maintain that accuracy throughout the range.

After having calibrated thousands over the years, even the best lab environmental monitors struggle to achieve 0.5 deg F accuracy over more than one point. If you want better than that, you are looking at higher end platinum resistance thermometers with polynomial curve fits to correct for non-linearity.

For what it is worth, I have taken my Ecowitt sensors into the lab and calibrated them to troubleshoot a humidity issue I was having with Ecobee. Yes, they read to 0.1. NO, they are not that accurate. The best multipoint linearity I got out of them was 0.7. But, I was able to make a band around 70F read a lot better. As to the humidity, you are looking at real world +/- 5% over the range. (which is actually on par with top end lab monitors)

2 Likes
10

100% this.

2 Likes
11

Nothing made by Ecowitt is going to be remotely accurate. I threw away their weather station.

12

You missed my comments about the actual values I was able to get from my 5 ecowitt sensors. Mine are adequately accurate (within 0.7 degrees) throughout the entirety of their range. However, I keep in perspective that they are in a house that doesn't have laminar flow throughout, with positive pressure, to keep a uniform temperature throughout the house. Even then, the example I gave of a dimensional calibration lab can only keep temperatures uniform and constant to +/- 0.5 degrees (ALL thermostats have swing temperatures). And that is usually relegated to a "box within a box" design where you have double door entry and the lab is usually centrally located within the center of a lab where the external rooms are also held at constant temperatures.

My real point was not to be argumentative. It was more to be informative around setting real expectations around what you will get from consumer grade equipment. The OP was looking for a thermostat with 0.1 deg F resolution (I suspect they really meant sensitivity), when that resolution has little to do with controlling what the temperature would actually be in the space they are trying to condition. They could get the same results by just reducing the deadband as far as they can, and then offsetting the temperature reading as much as they need to in order to get the results they need at the specific location they needed to control.

The best thermostats I have seen at consumer grade have a minimum 0.5 Deg F deadband. Which means that the temp will swing a full degree between highs and lows. This is usually a protection for the HVAC equipment itself to prevent shortcycling the compressor. The only use case I have EVER seen (I am sure that there are others) for that kind of temperature control is where tight length measurements are being made (We work to uncertainties of ~2-5 microinches per inch), and that allows the temp to swing a full half degree. We also do not rely on the wall thermostat or room temperature monitors. We measure the temperature at the actual part using equipment that is known to be accurate enough. That temperature is ALWAYS different than what the wall temperature says.

2 Likes
13

I agree. I was using the main sensor on the gateway for my living room heat for awhile. I noticed when graphing them they are a bit jumpy, and they jump to discrete F increments.

Here are my current outside temps from Ecowitt for the last 1/2 hour:

Summary

48.02
47.84
47.84
47.84
47.84
47.84
47.66
47.66
47.66
47.66
47.66
47.66
47.48
47.48
47.48
47.66
47.66
47.48
47.48
47.48
47.48
47.48
47.48
47.3
47.3
47.3
47.3
47.3
47.12
47.12
47.12
47.12

So roughly they resolve to only about .2 degrees, regardless of the hundredths being reported, as it is rounding from .1 C accuracy. Unlike the T6s that appear to round from 1 C.

My Zigbee Sensors report in hundredths, but they do not repeat the same .1 C rounding values like the Ecowitt. Since I am tracking rise and fall to anticipate when to turn on the furnace, it is good that they are consistent when rising and falling, and they don't jump much. I usually get three or four readings between a degree, depending when they report. I force them to report on small changes using kkossev's driver.

Here is one of my Zigbee Sensor temps, over the last several hours:

Summary

68.23
68.23
68.23
68.23
67.66
67.66
67.66
68.22
68.22
68.13
68.13
68.13
67.68
67.68
68.27
68.27
68.27
68.13
68.13
68.02

I just realized I am still plotting T6 Zwave Temps in that room. For a laugh, here are the T6 Zwave Temps compared to the Zigbee Sensor. T6 on Left:

Summary
69.5 68.34
70.5 68.34
69.5 68.34
70.5 68.34
69.5 68.34
70 68.34
69 68.34
70 68.34
69 68.34
70 68.34
69 67.78
69 67.78
69 67.78
67 67.98
67 67.98
70 67.98
69 67.98
70 67.98
69 67.98
70.5 67.98
68.5 67.98
68.5 67.98
68.5 67.98

It is calibrated different, but beyond that all the T6 does is jump either to a .5 change, or a 1.0 change. How it makes that determination is beyond me, it may be tracking the last temp seen to calculate a two point running average or something, because it is not a C conversion with Zwave temps on.

I don't know how you get those kinds of details about a thermostat until you buy it and try it.

Edit: Ah, I forgot that those T6 temps are being manipulated by my thermostat controller, so that the T6 on the wall will read what the Zigbee temp sensor rounds to. So ignore those, that is my app trying to hone-in on the display temp based on feedback from the thermostat after changing the calibration number using the command in the T6 driver.

14

The issue is actually pretty complicated. Due to the thermal mass of the slab there is a major lag between input power and thermal result in air temperature. This is a classical control theory problem and is inherently unstable. The algorithms to solve these are typically proportional-integral-derivative types. Much more involved than a simple set value and hysteresis band.

I suspect some of the better name brand thermostats employ such algorithms but I don't know which ones. Asking for 0.1 degree accuracy will cause a lot of cycling so you better use a solid state relay instead of a contact type as mechanical contacts will be pitted and destroyed in no time. I believe you are asking for precision beyond what such a system can reasonably deliver. I don't know many people who can discern 0.1 degree temp delta. Not many can even discern 1 degree. Its like the carpenter who says the 1/32 lines on the ruler are only useful for filing your fingernails.

2 Likes
15

I think what @Madcodger was asking came down to temperature reporting, not a degree of accuracy from the system.

How the thermostat acts on those increments of .1 is an entirely different matter, and also important, as you point out.

1 Like
16

Changing with a one degree swing, I don't notice. I do notice a two degree swing, like between 69 and 71 when the thermo is set for 70. Most thermostats that measure temp to one degree, end up with 3-5 degree swings around setpoint due to time to get heat and time to stop heating. Display temp on the screen is manipulated, I've found, at least on the T6. Most of the time it reads the setpoint, rarely will you see it go one degree over/under, even when real delta is two or more degrees. Partly due to how bad the sensor accuracy is.

17

With your objective of making the system easy to understand and service for someone other than yourself, do you think you have exhausted the search for system controls for your type of heating? Obviously, the system needs pid control. Most of the thermostat systems for radiant floors have this functionality. I just wonder if the best solution is using a purpose built off tge shelf system that works. Perhaps there is something out there now that gives you the features you want and the usability people may want when you aren't around to care for it.

I know a couple of old timers with radiant heat that swear by using old school murcury thermostats. The looks does go with their mid century houses.

18

You are correct, @chrisbvt - it's the reporting at 0.1 degree F (or at least less than 1 degree F) that is the real key here. Thanks for your responses.

I really just wondered if anyone knew of a thermostat that could do that, to be inserted into my system (the one that actually works pretty darn well).

Mods, feel free to close this thread (or just delete it).

19

I think many in floor heating systems rely on an in slab thermo probe and an air temp probe. Then the tstat had logic based on both of those inputs. The control challenge is the thermal mass of the slab dampens the turn on response but once air probe is satisfied the residual heat in the slab overshoots the set point.

I suspect these dual probe tstats have algorithms that predict a reaching of set temp and shut the heat off before temp is satisfied such that the cooling off of the slab achieves the set point.

1 Like
20

Why should the thread be closed or deleted?

It’s been an interesting and informative discussion so far.