A smart thermostat is not a good choice with a (Mitsubishi) heat pump

A month or so ago, we replaced our old forced-air heating and cooling system with a heat pump-based system. Since we have ducting, floor registers, etc., we chose a heat pump-based system with a central air handler, rather than a mini-split solution. The system chosen was a Mitsubishi Electric MEL SVZ-KP36NA, and what I describe here is specific to that system, although I strongly suspect it is true for all Mitsubishi systems, even mini-splits, and, quite possibly, other heat pump systems.

This post will not go into the details of the why and how of converting to a heat pump-based solution, but will discuss my experience with using an ecobee smart thermostat to control this system. I will explain why this is not a good choice. Much of this discussion applies to any brand smart thermostat that is wired into the system to replace a traditional 24V thermostat, so even though I discuss ecobee here, the problems are not unique to it.

About heat pump heating and cooling

I will only discuss heating in this post, but by and large, the same issues apply to cooling (though sometimes in the opposite temperature direction). A heat pump moves (pumps) heat from one location to another. To heat your house, a heat pump uses electricity to extract heat from outside air and circulate it through your ducts or a mini-split inside your house. In this example, you can think of a refrigerator working in reverse: it heats what’s inside (your house), at the expense of cooling what’s outside.

This is a very energy-efficient process, but unlike forced-air gas furnaces, it does not blow hot air around. The heating process is more gradual, and air temperatures involved are lower, although still warmer than the ambient temperature, of course. Heat pump systems work best and most efficiently if they are always on and maintain a fairly constant temperature.

By carefully monitoring the current temperature and the desired temperature (set point), algorithms control the compressor (the unit outside) speed and the air handler fan speed to reach and maintain heat efficiently, and, generally, at a low fan speed to minimize noise.

Traditional 24V thermostats

For a long time, all kinds of thermostats have been used to control heating systems. They generally operate at a supply voltage of 24V and, in essence, open or close a circuit back to the furnace, depending on the presence or absence of heating demand.

Heating demand occurs whenever the ambient temperature sensed by the thermostat is below the user-selected setpoint temperature. Old-style thermostats do this in a very binary manner: the circuit closes (activates) and opens (de-activates) the heat demand signal. Electronics in the furnace may respond directly by switching the furnace on or off, or may have a little more sophistication (which I won’t get into) to optimize behavior, but that is essentially how such a thermostat works.

In practical scenarios, there may be a second circuit to allow a thermostat to control first-stage heating and second-stage heating. It would then use the difference between ambient and setpoint temperatures to decide whether to activate only the first stage or both the first and second stages of heating, with the latter presumed to have greater heating capacity. The stages could correspond to more burners activating and/or higher fan speeds, or to firing up another heat source, such as electric heating, an oil-based furnace, etc.

ecobee

An ecobee smart thermostat has all kinds of smarts, but is designed to replace a traditional 24V thermostat. Because it is a replacement, it generally has the option to wire control of stage 1 heating, stage 2 heating, and sometimes fan speed control (and ditto for cooling). Yet all it can do is tell the furnace to switch to stage 1 or 2 heating by closing the circuit. The furnace will not know the current temperature or the setpoint.

Besides the fancy display, remote access features from your phone, etc., the “smart” part comes from ecobee observing your system’s response to actual heating scenarios and temperature differences, and then offering to start heating before a time where a certain setpoint is scheduled so that the temperature is achieved at that time, rather than starting a heating effort at that time. Of course, there is more to it, but that is the basic idea. For traditional heating systems, a smart thermostat can be an enormous improvement, both in terms of usability, comfort, and sometimes fuel economy.

What an ecobee (or other smart thermostat meant as a 24V replacement) can not do is communicate with the furnace in a more detailed manner. That requires a “communicating thermostat.”

Communicating thermostat

A communicating thermostat will, regularly, communicate to the furnace electronics the current ambient temperature and the setpoint temperature. It is then up to the furnace electronics to control the compressor and fans to deliver what it thinks is an optimal experience, based on user comfort and system capabilities. Of course, such thermostats may also have scheduling functions, displays, etc., but in essence, it comes down to just the current and setpoint temperatures.

Many heatpump systems, including the Mitsubishi I had installed, are designed around this model of a communicating thermostat and cannot (directly) be operated by a traditional thermostat. Instead of supplying 24V and various wires to close circuits for stage 1 and 2 heat, etc., these thermostats usually have just a few wires to use some kind of serial communications protocol. Digital temperature information travels along these wires, and the furnace activates the compressor and fans as it sees fit, based on the information it receives, some installer settings, and, in some cases, what it has learned about your home’s response to heating or cooling. Oftentimes, they also incorporate outside temperature into the decision-making.

Bridging old and new worlds

If, for some reason, you need to or want to use a 24V type thermostat with a system that really requires a communicating thermostat, some solution is needed that makes the system think it has a communicating thermostat. In the case of Mitsubishi, that comes in the form of a Mitsubishi 3rd Party Thermostat Interface PAC-US445CN1.

This is a small box with electronics and connectors for traditional 24V wires (R/Rc/Rh for 24V power, W/W1/W2 - heating demand, Y/Y1/Y2 - cooling demand, G - fan, C - Common, others), as well as a connection to the furnace using the manufacturer’s protocol. In the case of Mitsubishi, that is a serial protocol using CN105 connectors.

The box’s function is to take the demand signals from the traditional thermostat and, somehow, simulate a communicating thermostat. Computer algorithms determine exactly how this is done, and we won’t go into the precise details here, but suffice it to say that this is only as functional as the quality of the algorithm. It is also limited by the available information (other sensors).

An example: Upon receiving stage 1 heat demand (W1 wire), the PAC-US445CN1 will act as if the communicating thermostat has a set point 2F higher than the current temperature (as indicated by the sensor inside the air handler). This causes the unit to start producing warm air and blowing it around. As long as W1 remains energized (which will be the case as long as the 24V thermostat thinks its own set point temperature has not been reached), the PAC re-evaluates. If

• The internal temperature has gone down; it thinks things are cooling faster than the heating is doing its work. It raises the simulated set point by another 2F, under the assumption that a greater difference will cause the unit to produce more heating capacity
• The internal temperature has gone up by up to 2°F. It assumes things are working, but we’re not there yet (W1 still energized), so it adds another 1F to the simulated set point.
• The internal temperature has gone up by exactly 3F (which is a little weird), and it does nothing
• The internal temperature has gone up by more than 3°F. It assumes things are heating up (too) fast, and it lowers the set point by 1F.

The above is a simplified description, since other settings can affect behavior, and I only discuss stage 1 demand; stage 2 and cooling are also handled.

Does it work?

The algorithm described above, in theory, works, and it also works in practice in the sense that ultimately the 24V thermostat will see its set point reached or exceeded, it will then de-energize the W1 wire, and heating operations will be wound down (the unit may keep the fan blowing for a little while longer). However, the lack of direct feedback on actual temperature changes and the primitive nature of the adjustment algorithm may, in practice, result in very undesirable behavior.

In my particular case, I observed that the system would spin up the outside compressor to maximum capacity very quickly and blow the air handler fan at maximum capacity. The result would typically be high energy consumption, lots of noise, and the temperature would overshoot the setpoint by several degrees.

Now that I have an installation with an actual communicating thermostat, the behavior is very different: in most situations, the compressor speed is much lower, and the fan speed is lower, but the system remains on longer. Overall energy consumption is substantially lower, noise levels are more comfortable, and temperatures swing less than before.

Why do people still install smart thermostats with heat pumps?

From my experience, customers want the perceived or real conveniences of a smart thermostat with their heating and cooling solution. As it so happens, there are few, if any, smart thermostats designed to work as communicating thermostats with various heat pump systems. So there is a dilemma. Many installers, including mine, are not sufficiently aware of the technology, and its limitations and benefits, and will “please” the customer by installing a smart thermostat and necessary adapter. They do not inform the customer of the downsides (perhaps because they don’t even know themselves), or alternative solutions.

One of the benefits customers seek in a smart thermostat is the ability to control heating and cooling remotely (from their phone, for example). Traditional 24V thermostats and most, if not all, standard communicating thermostats do not offer this capability. Smart thermostats do, but as described, this comes at a cost.

Mitsubishi offers something called Kumo Cloud with their communicating thermostat. While, in theory, this accomplishes both the goal of a communicating thermostat and remote control, scheduling, etc., this particular solution has been widely described as a horrible implementation to be avoided.

I did avoid it and went in search of other solutions, which I continued operating the ecobee with an adapter (out of necessity). I’ll write about the alternatives separately, later.