A simple electromechanical add-on was the solution to a problem with this classic thermostat of home heating systems.
I recently helped someone upgrade their heating-system control from the classic Honeywell T-86 thermostat, often called as “the Round” (Figure 1) to a newer, smart one. Doing this was literally an upgrade of several orders of magnitude and decades, as the existing one had been there for at least 30, and maybe 40 or 50 years.
A brief online search revealed that this thermostat was introduced in 1953; I couldn’t find when it was formally discontinued, but it wasn’t that long ago. Millions were sold worldwide and many of them are still in use. The basic design is so ingrained that Honeywell offers an electronic unit with the same fit and form factor, for those who like the simplicity (Figure 2). That replacement has a long-life Lithium battery which should be good for 10-20 years. One of the original T-86 units is in the classic design collection at Cooper Hewitt, the Smithsonian Design Museum.
The objective of this thermostat is simple and reliable: to provide a switch closure between two leads (a “hot” one, usually at 24 VAC, and a common) to call for heat when the ambient sensed temperature drops below the setpoint threshold; and to separate those leads when it goes above the that setpoint. What’s interesting is that it doesn’t use any electronics or electrical power; instead, it uses a spiraled bimetallic strip which expands and contracts in length and thus curls/uncurls with temperature increases or decreases. There is a sealed glass vial at the end of the curled strip with the two leads inserted through the glass and a tiny amount of mercury inside; this mercury makes and breaks the electrical connection between the two wires connection as the strip curls/uncurls and the vial’s orientation changes.
The reliability of this unit is a consequence of its operating principle and design and is proven by its long-term performance in the field. The gentle curling of the bimetallic strip does not induce metal fatigue for a long time, and the electrical contacts are sealed in the vial and so do not corrode. (Yes, the presence of mercury is a hazard if the ampule breaks, and it is a disposal issue as well.)
Still, the designers of this very effective solution to temperature control faced the same problem that all on-off (also called “two-position” or “bang-bang”), non-proportional control systems face: the tradeoff between the accuracy band around the setpoint versus on-off cycling rate (Figure 2). Faster on/off cycling around the setpoint results in a tight window but that frequent cycling (chattering) is annoying and causes excessive wear on the heating system elements (fans, blowers, ignitors, electrical heaters, switches, relays, and other components). The cycling rate is a function of the heating system output power and type (heated air, electrical heat, or hot-water radiators) and the house’s thermal mass and the associated thermal time constant, all of which are totally outside of the thermostat’s domain.
The problem is well known, as is the first step to a solution: add a little hysteresis to the loop (Figure 3). In this way, the thermostat has a deadband defined by turn-on and turn-off thresholds; it is typically set at around one degree. But there’s a problem: how do you actually build this into a non-electronic device? There’s no software, no analog op amps (it’s very easy to add hysteresis to an op amp), or other electronics, in fact, it’s an all-mechanical system switching a 24 VAC line.
The solution they implemented was simple and clever and had been used in other control units, but never on a long-life, mass-market, low-cost product. They added an anticipator (first used in temperature controllers in 1924), which is a tiny heater that warms the thermostat to “fool it” into thinking the temperature is higher than it really is. This causes the thermostat to turn the heating system off a “little bit” before the room has actually reached the setpoint, thus preventing overheating. The heater is a tiny nichrome wire wound on an insulator, self-powered by the 24 VAC of the system’s leads (a sort of energy harvesting) and positioned so it warms the bimetallic strip (Figure 4).
There are two challenges to this scheme: first, the nominal 24 VAC of the system is actually far lower, down to 15 or 12 VAC; thus, reducing the heater output. Second, the rise/fall time of the heating system is a function of its type. Hot water has the largest thermal mass and carry-though “momentum,” while forced hot air has less, and electric heat has the least.
To overcome these problems, the anticipate heater has a tiny, adjustable slider on the nichrome wire. This sets the amount of current passed through the heater: less resistance means more current, and thus more heat, by our well-known relationship P = I2R. The vendor even has some suggested settings, (Figure 5); the setting number refer to the anticipator current (in amps).
Although the anticipator lacks the flexibility of our modern, software-controlled heating profiles, which add time of day and other programming options, it’s quite effective at doing one thing, doing it well, doing it inexpensively, and doing it reliably. Of course, this heating element adds to complexity and decreases long-term reliability, but the unit still works even with the common open-failure mode even if the anticipator function is lost. That’s a much better failure mode than a thermostat getting locked in fully on or off mode.
Have you even resorted to a simple, non-software solution to a problem, even if a microcontroller was conveniently available? (I certainly have, see “How an electromechanical relay literally wrestled my software to the ground.”) Was it a tough “sell” to the project team and marketing folks? Or did people acknowledge that it was the solution that actually made the most sense?