I've noticed that the heater/pump on my inflatable hot-tub has a very "sloppy" thermostat. If I set it to maintain at, say, 37 degrees (C), it heats to 39, then goes off, until 35, when it comes back on again. It has a digital display showing the temperature, so this is not an approximation, it is literally flipping on/off immediately when the display gets to the temperature 2 degrees either side of my selection.

For a long time, I assumed that they have done this to minimise the number off On/Off cycles of the heating element, in order to prolong its life, but after a bit of searching, it seems that the life of electric heating elements is dictated by the On time, not the number of On/Off cycles.

So, presuming that 1) the heater is under microprocessor control, and 2) we know the temperature accurately and digitally, why would they choose to have so much "slop" in the way they maintain the temperature? They could easily choose to maintain it within ONE degree of the selection, and, personally, I would even have, say, a 10% duty cycle when it is AT the selected temperature.

  • 7
    \$\begingroup\$ But what if the heater is not under MCU control? And even if it is, the relay used to switch it has also electrical life of some amount of switching so the less you switch the less it needs replacing. Assuming it uses a relay. Semiconductors could be used but if the heater is too high power then a relay must be used. Basically, you are asking why an electrical device is made to work in a certain way and only manufacturer can know it. \$\endgroup\$
    – Justme
    Commented Mar 16, 2022 at 11:48
  • 10
    \$\begingroup\$ The life of the heating element isn't determined by the number of cycles, but the life of the switch often is. \$\endgroup\$
    – user16324
    Commented Mar 16, 2022 at 12:22
  • 1
    \$\begingroup\$ @SolomonSlow Interesting - well, on my one the pump runs all the time and the heater goes on/off separately, so I have always assumed they just use a heating element like one from a kettle...? \$\endgroup\$
    – Lefty
    Commented Mar 16, 2022 at 14:42
  • 3
    \$\begingroup\$ If it's exactly +/- 2° then I'd say that's a pretty certain indication of MCU control. \$\endgroup\$
    – Lundin
    Commented Mar 16, 2022 at 15:27
  • 1
    \$\begingroup\$ Yeah, it's probably using a bimetalic strip rather than an MCU... If it was under MCU control, it'd probably be using a PID controller and PWM to heat the element to the setpoint. If it is under MCU control... Then it's just really lazy programming. \$\endgroup\$ Commented Mar 17, 2022 at 16:09

3 Answers 3


This method is known as bang-bang control. Its advantage is that it is cheap, simple and reasonably effective. It can even be done without any electronics, just using a bimetal switch.

As to why they didn't set a tighter deadband, one can only venture guesses. Maybe the heater is switched by a relay that could wear out or make an annoying clicking noise. Maybe they just used the first value that came to mind and called it a day. Maybe they do use a bimetal switch and the hysteresis (slop) isn't under their control.

Note that there is a limit to how small you can make the "slop", because the system has delays: When you turn on the heater, the temperature sensor doesn't see an immediate rise, and conversely when you turn it off the temperature will keep rising for a while. So maybe two degrees is as good as it gets.

If that's the reason, then they could have gone with a PID controller (your duty cycle idea) for a tighter regulation. However these need to be carefully tuned or they can also oscillate around the desired setpoint. Maybe it didn't seem worth the effort to them.

  • 2
    \$\begingroup\$ Great answer, thank you. I never considered the Relay as being a component vulnerable to excessive switching - but I bet that's the reason now you say it. I had also wondered if the delay might be an issue as well, but, in practice, the 800Kg of water are very good at smoothing the effect of the heater, so it goes to 39, then smoothly falls to 35. There is never any sense in which it oscillates between 2 temperatures, not even once or twice during a transition, it goes straight from one to the other, every time. \$\endgroup\$
    – Lefty
    Commented Mar 16, 2022 at 12:18
  • \$\begingroup\$ System delays were my first thought. Trying to sense faster than the approximate time to stability doesn't work too well. It's easy to see in a house when the system has been switched off for a while--it will switch off multiple times and then soon switch on again because the building has a lot more thermal mass than the air in it. \$\endgroup\$ Commented Mar 18, 2022 at 5:43

Perhaps the hysteresis is programmable through some kind of service menu.

That's a fairly common way to program this kind of controller. If it lost its memory thorough one or another scenario, +/-2°C might be the default or perhaps the limit (blank memories tend to have all 1's).

There is reason to have a reasonably large deadband in order that the relay life is acceptable, so they would not have it set to something silly like 0.1°C.

  • 2
    \$\begingroup\$ The water won't be a uniform temperature anyway (point heating, distributed heat loss) so depending on the exact flow, people getting in and out or just moving around will cause the pump to draw warmer/cooler water from different areas. That means there's little point trying to maintain tight control even if the relay had infinite life 0.5--1°C might be doable (and in the OP's case a ±1°C deadband and a setpoint 1°C higher than current would keep the top end the same while improving the bottom end - that's probably close to a realistic optimum) \$\endgroup\$
    – Chris H
    Commented Mar 17, 2022 at 11:11
  • 1
    \$\begingroup\$ @ChrisH Mixing is pretty strong in a pool and doubly so in a hot tub. The water circulates pretty quickly as well and water has huge thermal inertia. I wouldn't expect a sensor on the return line would see any significant fluctuations from people stirring the water a bit more or less. \$\endgroup\$
    – J...
    Commented Mar 17, 2022 at 18:17
  • 1
    \$\begingroup\$ @J... You're right, it's a circular tub, there aren't really any nooks and crannies for pools of cold water to get trapped in. The presence of bodies in the water never seems to make any difference to the temperature reading. It always behaves exactly as I described before - falling in temperature smoothly between upper limit and lower before switching on again. \$\endgroup\$
    – Lefty
    Commented Mar 17, 2022 at 18:23
  • 1
    \$\begingroup\$ @J... I'm not thinking about stirring so much as changing (blocking) the way the pump stirs it \$\endgroup\$
    – Chris H
    Commented Mar 17, 2022 at 18:32
  • 2
    \$\begingroup\$ @ChrisH To me, a hot tub is pretty close to the CSTRs we always assumed in Chem. Eng. I think +/-0.1°C reading from the sensor might be achievable, but of course that doesn't mean that the water is isothermal, that the sensor reading accurately represents the water temperature even close to the sensor or that the error in the reading is even time-invariant. The spec for professionally processing C41 color negative film was in that range, and it was pretty difficult to achieve in practice, even in a water bath. My answer pretty is pretty much limited to the way the controller itself functions. \$\endgroup\$ Commented Mar 17, 2022 at 18:36

On/off control is perfectly acceptable. Yes it does tend to create an upper and lower limit oscillating cycle but 2° above and below the set-point just won't be a problem in this application. Maybe there is a cycle time of 1 minute with this level of resolution around the true set-point.

If you decided to make the resolution tighter (such as 0.5°) then the switching element would be operating about four times more quickly and, if it's a budget device then the switching element would wear-out 4 times more quickly.

Taking it further yes, you could enforce a PWM control system that uses a thyristor or MOSFET to control the element but then you'd need extra protection circuits against AC surges (especially during lightning storms) and this may be just too costly for the product.

  • 5
    \$\begingroup\$ My question was mainly out of pure curiosity, but, to some extent it was inspired by a very real sense in which a +/- 2 degree oscillation is not really acceptable in this application. The cycle time is not one minute, it is about 3 HOURS. In particular, if the water has fallen to 35 degrees just before I open the cover to use the hot-tub, the heater will not be able to keep-up with the continual loss of heat to the atmosphere and at 33 degrees it is no longer usable. Starting at 39, it falls to 35 in about an hour when uncovered, by which time it is too late to rectify the problem. \$\endgroup\$
    – Lefty
    Commented Mar 16, 2022 at 12:53
  • \$\begingroup\$ @Lefty: A well tuned PID controller could indeed do better in that case, by detecting the increased heat loss when you open the cover and immediately turning the heater on at full power to compensate for it. But it still can't solve the bigger problem, which is that your heater is apparently underpowered for your tub. If it was powerful enough to stop the temperature from dropping below 35 °C even with the cover open, you'd have much less of a problem even with the sloppy control. \$\endgroup\$ Commented Mar 16, 2022 at 23:40
  • \$\begingroup\$ @IlmariKaronen You're absolutely right, it needs more power to cope with the cover being removed - but the big selling point of this type of hot tub is that it plugs into a standard 15 Amp (240V) outlet, so they are very-much limited to an underpowered heater and reliant upon the buffering effect of the water to store heat. Even more reason to have more intelligent control of the heater in my opinion. But then they are also trying to sell them for £100s rather than £1000s so there are compromises all over it. \$\endgroup\$
    – Lefty
    Commented Mar 17, 2022 at 9:35
  • 1
    \$\begingroup\$ @Lefty If +/- 2 degrees C is really not acceptable to you, then perhaps your hot-tub has a setting to switch to Fahrenheit? If, so then if you are lucky, they use the same +/- 2 degrees, but in Fahrenheit. This will give you a tighter tolerance than in Celsius. \$\endgroup\$
    – Glen Yates
    Commented Mar 17, 2022 at 14:43
  • 1
    \$\begingroup\$ @Lefty it's 13A. But fuses blow so slowly that you could draw 15A indefinitely through one. Something else might overheat though. And breakers + cables are rated higher because they serve the whole ring main \$\endgroup\$
    – Chris H
    Commented Mar 17, 2022 at 18:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.