Hot Water Control Hysteresis?

Having some level of hysteresis is useful if there is some noise on the sensory data, particularly if you are trying to hover at a steady temperature in which case you may see quantisation noise. In practice the time lags due to the dynamics will usually mean that the setpoint is overshot and it takes quite a bit of time before the temperature then drops below this. So in a sense you will get natural filtering, but it does depend somewhat on where and how you are measuring the temperature and how well insulated your tank is, whether you see much quantisation noise. Anyhow hysteresis basically prevents switching the heat demand signal until the temperature has changed by the set hysteresis level relative to where it last switched. For example if the set point is 50C and the heat demand was switched off upon exceeding this then if hysteresis value was 3C it will not be switched back on until the temperature drops to below 47C. This works both ways. If you are seeing excessive noise then another approach might be to apply some filtering to the temperature signal. You can do this on the physical signal e.g using capacitance but this will not get rid of quantisation noise. Alternatively you could create a digitally filtered version of the raw temperature signal (using a logic filter object) and feed the filtered version of the temperature signal to the hot water controller. However filtering slows down the dynamics so one consequence will likely be bigger overshoots (though you might mitigate this a bit by adjusting the setpoint). As with most things it can sometimes be a tradeoff balance.

I personally have 3 UFH type thermistors on my own tank, buried through the insulating jacket such that they are touching the metal skin. I use the mid way one for the hot water controller though I have debated with myself whether it might be better to use the lower one or an average or some other permutation (that debate not yet concluded :-) ). I don't really see any noise affecting the signals in the control range - in other words the signal is either rising or falling steadily so no long steady stretches to expose quantisation noise. So actually I get by ok with 0 hysteresis, but probably a value of 1 would be a good start since the 8 bit quantisation of thermistor signals typically translates to a quantised resolution of ~0.4C.

So with a combination of turning the target temperature down a bit and a hysteresis of 3, I'm getting quite a consistent boiler run now, with no cycling. Thank you. Chris.

P.S I think logic filters are a bit beyond me......