temperature control circuit

Possibly a simple on/off, possibly one with a predictor -- but probably just on/off.

Assuming you're in North America I'd consider going to your local Home Depot or equivalent and getting a thermostat for a baseboard heater. They are designed to switch 120V loads (double check the current rating), and if you look you can probably find one that works decently.

Can't you just turn it down?

Possibly, although it may be lots of fun. PID controlled heaters need to be tuned to match the room that they're in, or you'll get bad juju ranging from overshoot to wild oscillations. If you want to do it for the challenge of it, either look on Ebay for industrial temperature controllers -- you may be able to luck out and find an old one for $20 or so. If you build one, read this first:

PID controlling a room heater is probably going to turn out like hunting the crumple-horned snorkack. It isn't a normal loop compensation problem, because the open-loop transfer function depends on a whole lot of things, e.g. doors opening or closing, drafts, and so on. These aren't just ordinary thermal forcing, because they change the functional relationship between what you care about (the rate of energy loss from your skin, basically) and the thermostat reading, and also change the coupling between room temperature and thermostat temperature.

Bang-bang control (i.e. a thermostat) hides most of these variables.

Cheers,

Phil Hobbs

Yea, I guess I did forget to mention that "the room you're in" changes when you open the door, or fill it full of relatives...

Bang-bang control is an exceptionally sensible approach, when you can stand the inevitable process variable cycling.

Usually the "actuators" (A/C unit and central heating) can't handle fast cycling and don't have proportional outputs, so the controller is not the limiting factor.

I'm sure there are metrology rooms that do have these things, but it's very expensive and uncommon.

Best regards, Spehro Pefhany

Actually, with a house full of relatives, I'd kinda tend to want to use the other kind of bang-bang control. ;-)

Cheers! Rich

Tim Wescott snipped-for-privacy@seemywebsite.com posted to sci.electronics.design:

Indeed. And when you consider that most typical (US) home heating and cooling processes are not throttleable either.

I have a Honeywell MagicStat3 thermostat installed, and according to the manual it uses a PI controller. It is connected to a natural gas fired furnace (no A/C), which heats up water, and pumps it around through radiators. The furnace output is regulated by a slow PWM (max 6 cycles per hour).

The PI parameters are not user-adjustable, but the systems works very well. Under normal circumstances, the temperature in the room never varies more than 0.5 oC. With a bang-bang control, the room temperature fluctuations would be uncomfortable, perhaps made worse by the fact that the radiator system has a large heat capacity.

I think that's nonsense. What exactly do you mean by "the temperature in the room"? Unless this is an interior room in a big building, it's very unlikely that the temperature differences across the room, or from floor to ceiling, or from air to walls, are as small as 0.5 C. When the radiator is on, there will be temperature differentials of tens of degrees inside the room--otherwise it'd take a *long* time to heat up.

I can possibly just believe that it holds the *thermostat* temperature to that level, assuming nobody ever opens a window, but that's because the thermostat is bolted to the wall, which has a huge thermal mass.

Cheers,

Phil Hobbs

What can we say? You know your relatives betetr than we do.

I was talking about the temperature fluctuations near the thermostat.

Obviously, the thermostat can only control its own temperature. The best it can do is keep itself at a constant temperature. If it can achieve that, we can play with all the parameters in the room to reach maximum comfort where we want it. When the temperature at the thermostat is already oscillating, there's a good chance that the oscillations will be even bigger in other places, such as near the radiators.

There are bigger temperature differences across the room. The air right above the radiators is quite warm, and my single pane windows are much colder. However, the radiators are all mounted just below the windows, such that the rising warm air acts as a very effective "air curtain" in front of the cooler windows. As a result, these hot and cold spots tend to be very localized.

As a test, I put a temperature logger in the living room yesterday. The attached Pt100 sensor was placed on a chest-high cabinet across the room from the thermostat, near a dining table a few feet away from a window. Here is the log:

The log starts in the morning when the room is still cold. It takes about half an hour to do the initial heating. Then there's a 1 degree overshoot, probably because the thermostat is mounted on the wall which keeps it cool for a while. It takes another 2.5 hours for the temperatures to equalize. After that, the rest of the day the temperature stayed between 19.8 and 20.3. The temperature wanders around a bit, as people move around (or when the kids leave the door open), but there's no sign of bang-bang oscillations.

How DARE you pollute the pure armchair-intellectual atmosphere of SED with actual data? ;)

Well, that does look like a pretty decent thermostat, it's true. OTOH it also doesn't look altogether like linear settling. Those asymmetrical triangles look like some sort of on-off effect to me--each time the heat comes on (or gets turned on harder, or whatever it is) there's just about exactly 0.5 degree overshoot, followed by a more-or-less linear slope of more-or-less repeatable slope, just as though the room was cooling down on its own.

Given that it's winter, I would expect rapid transients due to door opening to be negative, rather than positive, so it really does look as though the thermostat is doing some sort of bang-bang control, possibly with some automatic tuning of parameters.

Cheers,

Phil Hobbs