Better thermostat

[MODERATOR NOTE]: moved to a more appropriate forum.

I wanted to post in the thermostat section but it doesn't allow posting of new threads so I guess I post here.

The problem is how simply (and badly) the current crop of thermostats work. Most are really simple and, in the case of a furnace/heater, go on when the temperature drops below the setpoint, and off again when the temperature rises above the setpoint, with some hysteresis in-between. Some have a timer to prevent cycling too quickly (usually for the cooling function). OK so far.

I have a multi-stage heat pump and there the behavior of the thermostat becomes annoying. Here is an example:

The heating setpoint 70˚F. The first stage heating function doesn't come on until the temperature drops by 1˚F to 69˚F. If it is especially cold the temperature continues to drop until 68˚F, at which point the second stage kicks in. Now instead of maintaining the desired 70˚F, it is cycling between 68˚F and 69˚F. If the temperature drops further, I think by 4˚, then and only then does the auxiliary resistance heat kick in. I can attest to the fact that, while we are having a really bad cold snap here in Texas, I am getting to enjoy the wide range of temperatures in my house. The algorithm of using 1˚F change for each is just too large a hysteresis. What I REALLY want is either a PI or PID thermostat. If one existed and I could find it, especially if it had a interface that would work for Indigo, problem solved.

However, it might not exist and I'd have to make one. One could easily use Indigo to implement this if there was a good high-accuracy temperature sensor, probably on the order of 0.1˚C or 0.2˚F resolution. The necessary relays, i.e. reverse (heat/cool), first stage on, second stage on, aux heat on, is pretty easy (4 relays). As soon as temperature drops below the setpoint, first stage on. Get differential (rate of drop) and if negative, activate second stage. If differential is still negative, activate aux heat. At this point temperature is still close to the setpoint. As soon as the differential goes positive and temperature is approaching the setpoint again, switch off aux heat, then second stage, then first stage. What would then happen is that it would end up cycling between off and first stage, first and second stage, and second and aux-heat stage, while actually keeping the temperature at my setpoint.

So, I know where to get a relay module that Indigo can control (Request: more support of RS485 so I don't have to depend on Z-wave or Insteon) but I am looking for a much better stand-alone thermostat, or a good, high-accuracy and high-resolution temperature sensor.

BTW, the various multi-sensor modules are not useful because the temperature sensing is biased by internal heat generated by the components. It needs to be something with a free-standing calibrated sensor that truly senses ambient temperature and has a reasonable amount of thermal mass/inertia.

Anyone have any information and/or ideas?

Thank you.

Ecobee thermostats are pretty smart, and will do "anticipatory" heating or cooling. They have some other name for it, but that's what it does. You should look into them.

OTOH, I think your expectations of what's possible with a home heating system are unrealistic. There's a huge amount of lag in any home HVAC, and a large variability between heat input and temperature change. While you can certainly measure temps to 0.1F at any one location, how many sensors are you going to use? And what's your algorithm for combining those sensors to get an "actual" temperature for the input? I've had houses that vary 4-5F just between rooms.

Yes, there can be large differences between rooms. I am not trying to address that. The key is the use of Proportional, Integral, Differential (PID) control algorithms that deal with lag and rates, and integrated error signals that deal with the fixed offset that afflicts proportional systems. The goal is to decrease response time while also preventing overshoot, which often happens when you try to get too much gain in the feedback system to deal with that offset.

For what I am talking about it is a balance between the three. The easiest one to talk about it the differential loop. If I am heating, turn on one level of heating (first stages) but then the temperature is continuing to go down, it is clear you are not putting enough heat in even if the temperature has only changed by a fraction of a degree. You know already that you need the second stage and maybe you haven't even lost 1/2˚F. Kick on the second stage even before your temperature drops more. In this way the system can recognize the need for more heat long before there is a significant change in temperature.

I am not sure that the integral part of the PID loop is applicable because it works with accumulated error. It integrates the error and creates an increasing error signal over time, which corrects the proportional loop back to zero error. If I had one of the newer, continuously-variable-speed heat pumps then the Integral part of the loop would become very significant.

As for the differences between rooms, that can largely be dealt with by having optimized per-room return ducts. I have already done this and reduced that problem substantially.

One other thing I have considered is having sensors in every room so I can see how temperature varies per-room, which may lead to adding a zone, or changing which sensor the system is following, depending on who is in which room.

In any case, in my house, the temperature may vary as much as 4-5 degrees with the system switching between the various stages. I want to eliminate that huge error in the system. (It is proportional only with a granularity of 1˚F or more, depending on stage.)

I have been thinking of hacking the behavior of the current thermostat by using Indigo's ability to look at the thermostat's measured temperature and then moving the setpoints for the stages by changing the temperature setting of the thermostat. I put the desired set-point into a variable and initially load that into the thermostat. As soon as the temp drops by 1 degree, the thermostat switches to stage 1 and bumps the temp up by 1 degree. Now it will cycle between stage 1 and stage 2 without having to drop by 2 degrees. It is probably pretty easy to see how this would work but with a thermostat resolution of only 1˚F, it will be hard to keep the temp fluctuation at anything less than 1 degree. (It will actually be slightly more.) This is why I want a sensor with better resolution so I can detect the rate of change (differential) with much less than 1˚ of temp change, giving me a much tighter control loop.

BTW, I am sorry I didn't stop to visit. I had some airplane problems, and then had to scurry home.